DeepCell_tn_nuclear_K2c.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "330e05e3",
   "metadata": {},
   "source": [
    "**Further train DeepCell Models with K's Data and Make Predictions**<br>\n",
    "Starting from model trained with tn nuclear data, further train the model with resized K's data that is cut into 25x7 512x512 patches and masks on the lines are removed along with pixels identified as in the masks on the lines.<br>\n",
    "200 epochs, semanticdatagenerator, resized<br>\n",
    "Best mAP 0.48 at mpp=1.3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "a72e9c51",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T21:15:28.395635Z",
     "iopub.status.busy": "2022-11-05T21:15:28.394291Z",
     "iopub.status.idle": "2022-11-05T21:15:28.397091Z",
     "shell.execute_reply": "2022-11-05T21:15:28.397687Z"
    }
   },
   "outputs": [],
   "source": [
    "starting_model_path = 'tn1.0_nuclear_20221102.h5' # the model trained with nucleus data from the Tissuenet V1.0 dataset\n",
    "INPUT_PATH=\"images/training_resized_512x512_a/\"\n",
    "model_name = 'tn1.0_nuclear_K_512x512resized_a' \n",
    "model_path = '{}.h5'.format(model_name)\n",
    "epochs=200 # 20 sec per step "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "90f2183c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T21:15:28.406214Z",
     "iopub.status.busy": "2022-11-05T21:15:28.405447Z",
     "iopub.status.idle": "2022-11-05T21:15:35.440395Z",
     "shell.execute_reply": "2022-11-05T21:15:35.441407Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.7.1\n",
      "0.11.1\n"
     ]
    }
   ],
   "source": [
    "import syotil\n",
    "\n",
    "import numpy as np\n",
    "from skimage import io\n",
    "from matplotlib import pyplot as plt\n",
    "%matplotlib inline\n",
    "from timeit import default_timer\n",
    "import os\n",
    "from skimage import img_as_ubyte\n",
    "\n",
    "import deepcell\n",
    "from deepcell import image_generators\n",
    "from deepcell.utils.train_utils import rate_scheduler, get_callbacks, count_gpus\n",
    "from deepcell_toolbox.utils import resize, tile_image, untile_image\n",
    "from deepcell_toolbox.deep_watershed import deep_watershed\n",
    "from deepcell.losses import weighted_categorical_crossentropy\n",
    "from deepcell.model_zoo.panopticnet import PanopticNet\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.optimizers import SGD, Adam\n",
    "from tensorflow.keras.losses import MSE\n",
    "\n",
    "print(tf.__version__)\n",
    "print(deepcell.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "938b06af",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T21:15:35.448046Z",
     "iopub.status.busy": "2022-11-05T21:15:35.446864Z",
     "iopub.status.idle": "2022-11-05T21:15:35.452526Z",
     "shell.execute_reply": "2022-11-05T21:15:35.453288Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "175\n",
      "['images/training_resized_512x512_a/M872956_JML_Position8_CD8_train21_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train13_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train20_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train6_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train24_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train4_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train1_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train21_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train17_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train15_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train7_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train5_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train1_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train22_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train24_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train7_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train18_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train13_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train9_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train21_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train14_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train6_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train2_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train14_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train2_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train15_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train3_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train12_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train12_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train3_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train20_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train20_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train17_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train23_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train17_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train15_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train7_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train10_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train9_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train1_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train5_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train17_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train9_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train21_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train2_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train12_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train4_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train6_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train12_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train2_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train8_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train6_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train8_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train3_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train14_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train16_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train4_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train19_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train10_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train16_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train24_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train1_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train23_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train14_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train3_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train18_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train6_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train3_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train4_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train20_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train15_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train1_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train11_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train24_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train10_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train12_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train1_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train15_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train21_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train13_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train11_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train6_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train3_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train20_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train23_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train14_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train18_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train18_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train12_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train8_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train15_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train0_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train21_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train11_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train7_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train20_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train10_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train11_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train22_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train9_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train10_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train16_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train19_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train2_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train8_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train23_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train23_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train20_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train16_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train11_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train0_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train4_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train2_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train8_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train19_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train7_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train3_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train22_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train19_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train22_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train14_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train5_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train13_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train8_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train9_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train24_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train16_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train0_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train4_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train8_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train17_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train13_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train18_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train19_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train23_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train11_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train9_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train19_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train21_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train5_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train22_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train0_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train12_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train14_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train10_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train0_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train0_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train19_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train5_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train1_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train7_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train4_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train13_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train9_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train17_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train16_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train17_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train7_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train5_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train13_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD8_train16_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train23_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train10_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train24_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train5_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train18_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position13_CD3_train24_img.png', 'images/training_resized_512x512_a/M872956_JML_Position9_CD3_train0_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train2_img.png', 'images/training_resized_512x512_a/M926910_CFL_Position7_CD3_train22_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train11_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train18_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD3_train6_img.png', 'images/training_resized_512x512_a/M872956_JML_Position8_CD4_train15_img.png', 'images/training_resized_512x512_a/M872956_JML_Position10_CD3_train22_img.png']\n"
     ]
    }
   ],
   "source": [
    "import glob\n",
    "FILENAMES = glob.glob(INPUT_PATH+\"*_img.png\")\n",
    "print(len(FILENAMES))\n",
    "print(FILENAMES)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "1bac06f7",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T21:15:35.459462Z",
     "iopub.status.busy": "2022-11-05T21:15:35.458879Z",
     "iopub.status.idle": "2022-11-05T21:15:47.339030Z",
     "shell.execute_reply": "2022-11-05T21:15:47.339963Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-11-05 14:15:39.478663: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1525] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 14657 MB memory:  -> device: 0, name: Tesla V100-PCIE-16GB, pci bus id: 0000:04:00.0, compute capability: 7.0\n",
      "2022-11-05 14:15:39.480065: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1525] Created device /job:localhost/replica:0/task:0/device:GPU:1 with 14657 MB memory:  -> device: 1, name: Tesla V100-PCIE-16GB, pci bus id: 0000:06:00.0, compute capability: 7.0\n",
      "2022-11-05 14:15:39.481260: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1525] Created device /job:localhost/replica:0/task:0/device:GPU:2 with 14657 MB memory:  -> device: 2, name: Tesla V100-PCIE-16GB, pci bus id: 0000:08:00.0, compute capability: 7.0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(175, 512, 512, 1)\n",
      "(175, 512, 512, 1)\n"
     ]
    }
   ],
   "source": [
    "imgs = [io.imread(CURR_IM_NAME) for CURR_IM_NAME in FILENAMES]\n",
    "X_train = tf.stack(imgs)\n",
    "X_train = np.expand_dims(X_train, axis=-1)\n",
    "print(X_train.shape)\n",
    "\n",
    "masks = [io.imread(CURR_IM_NAME.replace(\"img\",\"masks\")) for CURR_IM_NAME in FILENAMES]\n",
    "y_train = tf.stack(masks)\n",
    "y_train = np.expand_dims(y_train, axis=-1)\n",
    "print(y_train.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "5c1212c8",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T21:15:47.345273Z",
     "iopub.status.busy": "2022-11-05T21:15:47.344239Z",
     "iopub.status.idle": "2022-11-05T21:15:47.405031Z",
     "shell.execute_reply": "2022-11-05T21:15:47.404287Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2688"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.max(masks) # sanity check"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "73d0c5d8",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T21:15:47.412103Z",
     "iopub.status.busy": "2022-11-05T21:15:47.411278Z",
     "iopub.status.idle": "2022-11-05T21:15:48.030300Z",
     "shell.execute_reply": "2022-11-05T21:15:48.031163Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Removing 10 of 175 images with fewer than 5 objects.\n",
      "302812\n"
     ]
    }
   ],
   "source": [
    "# create dataloader\n",
    "seed=0\n",
    "min_objects = 5\n",
    "\n",
    "transforms = ['inner-distance', 'outer-distance', 'fgbg']\n",
    "transforms_kwargs = {'outer-distance': {'erosion_width': 0}}\n",
    "\n",
    "# use augmentation for training but not validation\n",
    "datagen = image_generators.SemanticDataGenerator(\n",
    "    rotation_range=180,\n",
    "    fill_mode='reflect',\n",
    "    zoom_range=(0.75, 1.25),\n",
    "    horizontal_flip=True,\n",
    "    vertical_flip=True)\n",
    "\n",
    "datagen_val = image_generators.CroppingDataGenerator()\n",
    "\n",
    "batch_size = 4 # 8 causes memory outage\n",
    "\n",
    "train_data = datagen.flow(\n",
    "    {'X': X_train, 'y': y_train},\n",
    "    seed=seed,\n",
    "    transforms=transforms,\n",
    "    transforms_kwargs=transforms_kwargs,\n",
    "    min_objects=min_objects,\n",
    "    batch_size=batch_size)\n",
    "\n",
    "val_data = None\n",
    "# datagen_val.flow(\n",
    "#     {'X': X_val, 'y': y_val},\n",
    "#     seed=seed,\n",
    "#     transforms=transforms,\n",
    "#     transforms_kwargs=transforms_kwargs,\n",
    "#     min_objects=min_objects,\n",
    "#     batch_size=batch_size)\n",
    "\n",
    "# get number of training and validation instances\n",
    "\n",
    "cnts_train = [np.max(y_train[i,...]) for i in range(y_train.shape[0])]\n",
    "print(np.sum(cnts_train)) # total number of training instances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "818ee859",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T21:15:48.037349Z",
     "iopub.status.busy": "2022-11-05T21:15:48.036280Z",
     "iopub.status.idle": "2022-11-05T21:15:48.439435Z",
     "shell.execute_reply": "2022-11-05T21:15:48.440369Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[  1   1   1   1   2   4   4   4   5   5   6   6   7   7   8   8   9   9\n",
      "  11  13  13  14  15  15  16  16  17  18  19  20  20  21  22  23  23  25\n",
      "  25  26  26  27  29  29  30  32  33  33  34  35  36  38  38  39  39  39\n",
      "  39  40  41  41  42  43  43  44  44  44  44  45  45  48  49  49  49  49\n",
      "  50  51  51  52  52  55  57  58  58  59  59  59  61  61  63  65  65  66\n",
      "  66  68  68  70  70  72  72  74  75  76  77  78  79  79  80  81  82  82\n",
      "  82  83  84  84  85  88  88  89  89  89  89  89  92  94  95  95  98  99\n",
      " 101 101 102 102 102 102 102 103 103 104 104 106 109 114 115 115 115 115\n",
      " 118 120 120 123 124 125 127 128 129 130 132 135 136 143 143 144 144 144\n",
      " 146 147 149 153 153 160 161 163 165 171 178 178 182]\n",
      "[42, 42, 67, 143, 151, 120, 18, 128, 136]\n"
     ]
    }
   ],
   "source": [
    "mask_cnt=[len(np.unique(x)) for x in masks]\n",
    "print(np.sort(mask_cnt))\n",
    "indices = [mask_cnt.index(1)]\n",
    "indices.extend([i for i, x in enumerate(mask_cnt) if x == 1])\n",
    "indices.extend([i for i, x in enumerate(mask_cnt) if x == 2])\n",
    "indices.extend([i for i, x in enumerate(mask_cnt) if x == 4])\n",
    "print(indices)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "edaeac9e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T21:15:48.447181Z",
     "iopub.status.busy": "2022-11-05T21:15:48.446055Z",
     "iopub.status.idle": "2022-11-05T21:15:48.953669Z",
     "shell.execute_reply": "2022-11-05T21:15:48.954548Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "136\n",
      "images/training_resized_512x512_a/M872956_JML_Position8_CD4_train9_img.png\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/app/software/scikit-image/0.19.1-foss-2021b/lib/python3.9/site-packages/skimage/io/_plugins/matplotlib_plugin.py:150: UserWarning: Low image data range; displaying image with stretched contrast.\n",
      "  lo, hi, cmap = _get_display_range(image)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[  0 194 539 899]\n",
      "[  0 194 539 899]\n"
     ]
    }
   ],
   "source": [
    "i=indices[-1]\n",
    "print(i)\n",
    "print(FILENAMES[i])\n",
    "# plt.subplot(1, 3, 1) \n",
    "io.imshow(io.imread(FILENAMES[i]))\n",
    "plt.show()\n",
    "# plt.subplot(1, 3, 2) \n",
    "io.imshow(io.imread(FILENAMES[i].replace(\"_img\",\"_masks\")))\n",
    "plt.show()\n",
    "# plt.subplot(1, 3, 3) \n",
    "tmp=syotil.masks_to_outlines(io.imread(FILENAMES[i].replace(\"_img\",\"_masks\")))\n",
    "io.imshow(tmp)\n",
    "plt.show()\n",
    "print(np.unique(masks[i]))\n",
    "print(np.unique(masks[i][1:510,1:510]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "0b79383a",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T21:15:48.976135Z",
     "iopub.status.busy": "2022-11-05T21:15:48.975000Z",
     "iopub.status.idle": "2022-11-05T21:15:49.279009Z",
     "shell.execute_reply": "2022-11-05T21:15:49.279946Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "i=5\n",
    "plt.subplot(1, 2, 1) # row 1, col 2 index 1\n",
    "io.imshow(X_train[i,:,:,0])\n",
    "plt.subplot(1, 2, 2) # row 1, col 2 index 1\n",
    "io.imshow(y_train[i,:,:,0])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9591736f",
   "metadata": {},
   "source": [
    "**The two cells below define and train the model.** They can be skipped if a trained model will be loaded."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "7ebc07f0",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T21:15:49.288319Z",
     "iopub.status.busy": "2022-11-05T21:15:49.287178Z",
     "iopub.status.idle": "2022-11-05T21:15:56.264143Z",
     "shell.execute_reply": "2022-11-05T21:15:56.265117Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-11-05 14:15:49.291567: W tensorflow/python/util/util.cc:368] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.\n",
      "/app/software/TensorFlow/2.7.1-foss-2021b-CUDA-11.4.1/lib/python3.9/site-packages/keras/optimizer_v2/adam.py:105: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
      "  super(Adam, self).__init__(name, **kwargs)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[('semantic_0', (None, 512, 512, 1)),\n",
       " ('semantic_1', (None, 512, 512, 1)),\n",
       " ('semantic_2', (None, 512, 512, 2))]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "semantic_classes = [1, 1, 2] # inner distance, outer distance, fgbg\n",
    "\n",
    "model = PanopticNet(\n",
    "    backbone='resnet50',\n",
    "    input_shape=(512,512,1),\n",
    "    norm_method='whole_image',\n",
    "    num_semantic_classes=semantic_classes)\n",
    "\n",
    "lr = 1e-4\n",
    "optimizer = Adam(lr=lr, clipnorm=0.001)\n",
    "lr_sched = rate_scheduler(lr=lr, decay=0.99)\n",
    "\n",
    "# Create a dictionary of losses for each semantic head\n",
    "\n",
    "def semantic_loss(n_classes):\n",
    "    def _semantic_loss(y_pred, y_true):\n",
    "        if n_classes > 1:\n",
    "            return 0.01 * weighted_categorical_crossentropy(\n",
    "                y_pred, y_true, n_classes=n_classes)\n",
    "        return MSE(y_pred, y_true)\n",
    "    return _semantic_loss\n",
    "\n",
    "loss = {}\n",
    "\n",
    "# Give losses for all of the semantic heads\n",
    "for layer in model.layers:\n",
    "    if layer.name.startswith('semantic_'):\n",
    "        n_classes = layer.output_shape[-1]\n",
    "        loss[layer.name] = semantic_loss(n_classes)\n",
    "        \n",
    "model.compile(loss=loss, optimizer=optimizer)\n",
    "\n",
    "model.load_weights(starting_model_path, by_name=True)\n",
    "\n",
    "[(layer.name, layer.output_shape) for layer in filter(lambda x: x.name.startswith('semantic_'), model.layers)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "b1fab4f5",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T21:15:56.285376Z",
     "iopub.status.busy": "2022-11-05T21:15:56.284197Z",
     "iopub.status.idle": "2022-11-05T22:30:36.311347Z",
     "shell.execute_reply": "2022-11-05T22:30:36.312395Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training on 3 GPUs.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-11-05 14:15:56.277817: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1525] Created device /device:GPU:0 with 14657 MB memory:  -> device: 0, name: Tesla V100-PCIE-16GB, pci bus id: 0000:04:00.0, compute capability: 7.0\n",
      "2022-11-05 14:15:56.278719: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1525] Created device /device:GPU:1 with 14657 MB memory:  -> device: 1, name: Tesla V100-PCIE-16GB, pci bus id: 0000:06:00.0, compute capability: 7.0\n",
      "2022-11-05 14:15:56.279600: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1525] Created device /device:GPU:2 with 14657 MB memory:  -> device: 2, name: Tesla V100-PCIE-16GB, pci bus id: 0000:08:00.0, compute capability: 7.0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-11-05 14:16:16.359256: I tensorflow/stream_executor/cuda/cuda_dnn.cc:366] Loaded cuDNN version 8202\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "41/41 [==============================] - ETA: 0s - loss: 0.0150 - semantic_0_loss: 0.0036 - semantic_1_loss: 0.0093 - semantic_2_loss: 0.0021WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 62s 620ms/step - loss: 0.0150 - semantic_0_loss: 0.0036 - semantic_1_loss: 0.0093 - semantic_2_loss: 0.0021 - lr: 1.0000e-04\n",
      "Epoch 2/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0127 - semantic_0_loss: 0.0031 - semantic_1_loss: 0.0081 - semantic_2_loss: 0.0015WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 531ms/step - loss: 0.0127 - semantic_0_loss: 0.0031 - semantic_1_loss: 0.0081 - semantic_2_loss: 0.0015 - lr: 9.9000e-05\n",
      "Epoch 3/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0121 - semantic_0_loss: 0.0029 - semantic_1_loss: 0.0077 - semantic_2_loss: 0.0015WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0121 - semantic_0_loss: 0.0029 - semantic_1_loss: 0.0077 - semantic_2_loss: 0.0015 - lr: 9.8010e-05\n",
      "Epoch 4/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0121 - semantic_0_loss: 0.0030 - semantic_1_loss: 0.0077 - semantic_2_loss: 0.0015WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 532ms/step - loss: 0.0121 - semantic_0_loss: 0.0030 - semantic_1_loss: 0.0077 - semantic_2_loss: 0.0015 - lr: 9.7030e-05\n",
      "Epoch 5/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0117 - semantic_0_loss: 0.0029 - semantic_1_loss: 0.0075 - semantic_2_loss: 0.0014WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0117 - semantic_0_loss: 0.0029 - semantic_1_loss: 0.0075 - semantic_2_loss: 0.0014 - lr: 9.6060e-05\n",
      "Epoch 6/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0119 - semantic_0_loss: 0.0029 - semantic_1_loss: 0.0076 - semantic_2_loss: 0.0014WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 524ms/step - loss: 0.0119 - semantic_0_loss: 0.0029 - semantic_1_loss: 0.0076 - semantic_2_loss: 0.0014 - lr: 9.5099e-05\n",
      "Epoch 7/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0117 - semantic_0_loss: 0.0029 - semantic_1_loss: 0.0074 - semantic_2_loss: 0.0014WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 530ms/step - loss: 0.0117 - semantic_0_loss: 0.0029 - semantic_1_loss: 0.0074 - semantic_2_loss: 0.0014 - lr: 9.4148e-05\n",
      "Epoch 8/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0118 - semantic_0_loss: 0.0029 - semantic_1_loss: 0.0075 - semantic_2_loss: 0.0014WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0118 - semantic_0_loss: 0.0029 - semantic_1_loss: 0.0075 - semantic_2_loss: 0.0014 - lr: 9.3207e-05\n",
      "Epoch 9/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0115 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0073 - semantic_2_loss: 0.0014WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0115 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0073 - semantic_2_loss: 0.0014 - lr: 9.2274e-05\n",
      "Epoch 10/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0118 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0075 - semantic_2_loss: 0.0014WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 537ms/step - loss: 0.0118 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0075 - semantic_2_loss: 0.0014 - lr: 9.1352e-05\n",
      "Epoch 11/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0114 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0072 - semantic_2_loss: 0.0014WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0114 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0072 - semantic_2_loss: 0.0014 - lr: 9.0438e-05\n",
      "Epoch 12/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0112 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0071 - semantic_2_loss: 0.0014WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 533ms/step - loss: 0.0112 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0071 - semantic_2_loss: 0.0014 - lr: 8.9534e-05\n",
      "Epoch 13/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0114 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0072 - semantic_2_loss: 0.0014WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0114 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0072 - semantic_2_loss: 0.0014 - lr: 8.8638e-05\n",
      "Epoch 14/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0112 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0071 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 537ms/step - loss: 0.0112 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0071 - semantic_2_loss: 0.0013 - lr: 8.7752e-05\n",
      "Epoch 15/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0114 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0072 - semantic_2_loss: 0.0014WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 537ms/step - loss: 0.0114 - semantic_0_loss: 0.0028 - semantic_1_loss: 0.0072 - semantic_2_loss: 0.0014 - lr: 8.6875e-05\n",
      "Epoch 16/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0111 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0071 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0111 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0071 - semantic_2_loss: 0.0013 - lr: 8.6006e-05\n",
      "Epoch 17/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0107 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0068 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 520ms/step - loss: 0.0107 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0068 - semantic_2_loss: 0.0013 - lr: 8.5146e-05\n",
      "Epoch 18/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0109 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0069 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0109 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0069 - semantic_2_loss: 0.0013 - lr: 8.4294e-05\n",
      "Epoch 19/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0109 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0069 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0109 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0069 - semantic_2_loss: 0.0013 - lr: 8.3451e-05\n",
      "Epoch 20/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0110 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0070 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 530ms/step - loss: 0.0110 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0070 - semantic_2_loss: 0.0013 - lr: 8.2617e-05\n",
      "Epoch 21/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0109 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0069 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 519ms/step - loss: 0.0109 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0069 - semantic_2_loss: 0.0013 - lr: 8.1791e-05\n",
      "Epoch 22/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0110 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0069 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0110 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0069 - semantic_2_loss: 0.0013 - lr: 8.0973e-05\n",
      "Epoch 23/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0108 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0068 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 21s 529ms/step - loss: 0.0108 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0068 - semantic_2_loss: 0.0013 - lr: 8.0163e-05\n",
      "Epoch 24/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0105 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0105 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0013 - lr: 7.9361e-05\n",
      "Epoch 25/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0107 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0068 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0107 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0068 - semantic_2_loss: 0.0013 - lr: 7.8568e-05\n",
      "Epoch 26/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0106 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0067 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0106 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0067 - semantic_2_loss: 0.0013 - lr: 7.7782e-05\n",
      "Epoch 27/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0107 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0067 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0107 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0067 - semantic_2_loss: 0.0013 - lr: 7.7004e-05\n",
      "Epoch 28/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0109 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0069 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 524ms/step - loss: 0.0109 - semantic_0_loss: 0.0027 - semantic_1_loss: 0.0069 - semantic_2_loss: 0.0013 - lr: 7.6234e-05\n",
      "Epoch 29/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0106 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0067 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0106 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0067 - semantic_2_loss: 0.0013 - lr: 7.5472e-05\n",
      "Epoch 30/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0104 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 538ms/step - loss: 0.0104 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0013 - lr: 7.4717e-05\n",
      "Epoch 31/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0102 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0102 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0013 - lr: 7.3970e-05\n",
      "Epoch 32/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0106 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0067 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0106 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0067 - semantic_2_loss: 0.0013 - lr: 7.3230e-05\n",
      "Epoch 33/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0104 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0104 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0013 - lr: 7.2498e-05\n",
      "Epoch 34/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0103 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 531ms/step - loss: 0.0103 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0013 - lr: 7.1773e-05\n",
      "Epoch 35/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0104 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 520ms/step - loss: 0.0104 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0013 - lr: 7.1055e-05\n",
      "Epoch 36/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0102 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0102 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0012 - lr: 7.0345e-05\n",
      "Epoch 37/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0106 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0067 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 524ms/step - loss: 0.0106 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0067 - semantic_2_loss: 0.0013 - lr: 6.9641e-05\n",
      "Epoch 38/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0102 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0102 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0013 - lr: 6.8945e-05\n",
      "Epoch 39/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0103 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 536ms/step - loss: 0.0103 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0012 - lr: 6.8255e-05\n",
      "Epoch 40/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0103 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 532ms/step - loss: 0.0103 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0013 - lr: 6.7573e-05\n",
      "Epoch 41/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0103 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 534ms/step - loss: 0.0103 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0012 - lr: 6.6897e-05\n",
      "Epoch 42/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0104 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0104 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0012 - lr: 6.6228e-05\n",
      "Epoch 43/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0103 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0013WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0103 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0013 - lr: 6.5566e-05\n",
      "Epoch 44/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0102 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0064 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 21s 518ms/step - loss: 0.0102 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0064 - semantic_2_loss: 0.0012 - lr: 6.4910e-05\n",
      "Epoch 45/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0101 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0101 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012 - lr: 6.4261e-05\n",
      "Epoch 46/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0104 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 524ms/step - loss: 0.0104 - semantic_0_loss: 0.0026 - semantic_1_loss: 0.0066 - semantic_2_loss: 0.0012 - lr: 6.3619e-05\n",
      "Epoch 47/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0101 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0101 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012 - lr: 6.2982e-05\n",
      "Epoch 48/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0102 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 522ms/step - loss: 0.0102 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0065 - semantic_2_loss: 0.0012 - lr: 6.2353e-05\n",
      "Epoch 49/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0098 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 520ms/step - loss: 0.0098 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012 - lr: 6.1729e-05\n",
      "Epoch 50/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0101 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0064 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0101 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0064 - semantic_2_loss: 0.0012 - lr: 6.1112e-05\n",
      "Epoch 51/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0099 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 522ms/step - loss: 0.0099 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012 - lr: 6.0501e-05\n",
      "Epoch 52/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0098 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 536ms/step - loss: 0.0098 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012 - lr: 5.9896e-05\n",
      "Epoch 53/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0101 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 541ms/step - loss: 0.0101 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012 - lr: 5.9297e-05\n",
      "Epoch 54/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0100 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0100 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012 - lr: 5.8704e-05\n",
      "Epoch 55/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0098 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 538ms/step - loss: 0.0098 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012 - lr: 5.8117e-05\n",
      "Epoch 56/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0101 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 533ms/step - loss: 0.0101 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012 - lr: 5.7535e-05\n",
      "Epoch 57/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0099 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 537ms/step - loss: 0.0099 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012 - lr: 5.6960e-05\n",
      "Epoch 58/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0099 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0099 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012 - lr: 5.6391e-05\n",
      "Epoch 59/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0099 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 531ms/step - loss: 0.0099 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012 - lr: 5.5827e-05\n",
      "Epoch 60/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0100 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 23s 543ms/step - loss: 0.0100 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0063 - semantic_2_loss: 0.0012 - lr: 5.5268e-05\n",
      "Epoch 61/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0099 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0099 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012 - lr: 5.4716e-05\n",
      "Epoch 62/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0097 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0061 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 520ms/step - loss: 0.0097 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0061 - semantic_2_loss: 0.0012 - lr: 5.4169e-05\n",
      "Epoch 63/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0097 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0061 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 537ms/step - loss: 0.0097 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0061 - semantic_2_loss: 0.0012 - lr: 5.3627e-05\n",
      "Epoch 64/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0099 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 524ms/step - loss: 0.0099 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012 - lr: 5.3091e-05\n",
      "Epoch 65/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0097 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0061 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0097 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0061 - semantic_2_loss: 0.0012 - lr: 5.2560e-05\n",
      "Epoch 66/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0096 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 21s 518ms/step - loss: 0.0096 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012 - lr: 5.2034e-05\n",
      "Epoch 67/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0097 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0061 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0097 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0061 - semantic_2_loss: 0.0012 - lr: 5.1514e-05\n",
      "Epoch 68/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0094 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 533ms/step - loss: 0.0094 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0012 - lr: 5.0999e-05\n",
      "Epoch 69/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0096 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 533ms/step - loss: 0.0096 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012 - lr: 5.0489e-05\n",
      "Epoch 70/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0096 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0096 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012 - lr: 4.9984e-05\n",
      "Epoch 71/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0098 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0098 - semantic_0_loss: 0.0025 - semantic_1_loss: 0.0062 - semantic_2_loss: 0.0012 - lr: 4.9484e-05\n",
      "Epoch 72/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0095 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0059 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 522ms/step - loss: 0.0095 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0059 - semantic_2_loss: 0.0012 - lr: 4.8989e-05\n",
      "Epoch 73/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0095 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 522ms/step - loss: 0.0095 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012 - lr: 4.8499e-05\n",
      "Epoch 74/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011 - lr: 4.8014e-05\n",
      "Epoch 75/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0095 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0059 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 520ms/step - loss: 0.0095 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0059 - semantic_2_loss: 0.0012 - lr: 4.7534e-05\n",
      "Epoch 76/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0095 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0095 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012 - lr: 4.7059e-05\n",
      "Epoch 77/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0095 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0059 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 519ms/step - loss: 0.0095 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0059 - semantic_2_loss: 0.0012 - lr: 4.6588e-05\n",
      "Epoch 78/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0093 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 21s 516ms/step - loss: 0.0093 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011 - lr: 4.6122e-05\n",
      "Epoch 79/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0093 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 524ms/step - loss: 0.0093 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0012 - lr: 4.5661e-05\n",
      "Epoch 80/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0096 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 23s 545ms/step - loss: 0.0096 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0060 - semantic_2_loss: 0.0012 - lr: 4.5204e-05\n",
      "Epoch 81/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0093 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0093 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0012 - lr: 4.4752e-05\n",
      "Epoch 82/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 531ms/step - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011 - lr: 4.4305e-05\n",
      "Epoch 83/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0094 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0059 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 533ms/step - loss: 0.0094 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0059 - semantic_2_loss: 0.0012 - lr: 4.3862e-05\n",
      "Epoch 84/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0093 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0093 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011 - lr: 4.3423e-05\n",
      "Epoch 85/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 535ms/step - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011 - lr: 4.2989e-05\n",
      "Epoch 86/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0012WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 531ms/step - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0012 - lr: 4.2559e-05\n",
      "Epoch 87/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 530ms/step - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011 - lr: 4.2133e-05\n",
      "Epoch 88/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 531ms/step - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 4.1712e-05\n",
      "Epoch 89/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 4.1295e-05\n",
      "Epoch 90/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011 - lr: 4.0882e-05\n",
      "Epoch 91/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 4.0473e-05\n",
      "Epoch 92/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 533ms/step - loss: 0.0093 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011 - lr: 4.0068e-05\n",
      "Epoch 93/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 3.9668e-05\n",
      "Epoch 94/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0058 - semantic_2_loss: 0.0011 - lr: 3.9271e-05\n",
      "Epoch 95/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 3.8878e-05\n",
      "Epoch 96/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 537ms/step - loss: 0.0092 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 3.8490e-05\n",
      "Epoch 97/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0092 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 3.8105e-05\n",
      "Epoch 98/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 530ms/step - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 3.7724e-05\n",
      "Epoch 99/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 520ms/step - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 3.7346e-05\n",
      "Epoch 100/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 3.6973e-05\n",
      "Epoch 101/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 531ms/step - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 3.6603e-05\n",
      "Epoch 102/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011 - lr: 3.6237e-05\n",
      "Epoch 103/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 536ms/step - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011 - lr: 3.5875e-05\n",
      "Epoch 104/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 530ms/step - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011 - lr: 3.5516e-05\n",
      "Epoch 105/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0092 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 3.5161e-05\n",
      "Epoch 106/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 3.4809e-05\n",
      "Epoch 107/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 3.4461e-05\n",
      "Epoch 108/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0092 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0092 - semantic_0_loss: 0.0024 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 3.4117e-05\n",
      "Epoch 109/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 531ms/step - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011 - lr: 3.3775e-05\n",
      "Epoch 110/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 541ms/step - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 3.3438e-05\n",
      "Epoch 111/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011 - lr: 3.3103e-05\n",
      "Epoch 112/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 533ms/step - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0057 - semantic_2_loss: 0.0011 - lr: 3.2772e-05\n",
      "Epoch 113/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 3.2445e-05\n",
      "Epoch 114/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011 - lr: 3.2120e-05\n",
      "Epoch 115/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 3.1799e-05\n",
      "Epoch 116/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 534ms/step - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011 - lr: 3.1481e-05\n",
      "Epoch 117/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 533ms/step - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011 - lr: 3.1166e-05\n",
      "Epoch 118/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 537ms/step - loss: 0.0091 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011 - lr: 3.0854e-05\n",
      "Epoch 119/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 538ms/step - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 3.0546e-05\n",
      "Epoch 120/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 3.0240e-05\n",
      "Epoch 121/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0088 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 538ms/step - loss: 0.0088 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 2.9938e-05\n",
      "Epoch 122/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 524ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.9639e-05\n",
      "Epoch 123/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0088 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 522ms/step - loss: 0.0088 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.9342e-05\n",
      "Epoch 124/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 2.9049e-05\n",
      "Epoch 125/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 531ms/step - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 2.8758e-05\n",
      "Epoch 126/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.8471e-05\n",
      "Epoch 127/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 518ms/step - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.8186e-05\n",
      "Epoch 128/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 519ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.7904e-05\n",
      "Epoch 129/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0088 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0088 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.7625e-05\n",
      "Epoch 130/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0089 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 2.7349e-05\n",
      "Epoch 131/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 531ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.7075e-05\n",
      "Epoch 132/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.6805e-05\n",
      "Epoch 133/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.6537e-05\n",
      "Epoch 134/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 535ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.6271e-05\n",
      "Epoch 135/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.6009e-05\n",
      "Epoch 136/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 536ms/step - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 2.5748e-05\n",
      "Epoch 137/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.5491e-05\n",
      "Epoch 138/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.5236e-05\n",
      "Epoch 139/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 543ms/step - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.4984e-05\n",
      "Epoch 140/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0090 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0056 - semantic_2_loss: 0.0011 - lr: 2.4734e-05\n",
      "Epoch 141/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0088 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0055 - semantic_2_loss: 0.0011 - lr: 2.4487e-05\n",
      "Epoch 142/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.4242e-05\n",
      "Epoch 143/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 535ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.3999e-05\n",
      "Epoch 144/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.3759e-05\n",
      "Epoch 145/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 524ms/step - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.3522e-05\n",
      "Epoch 146/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.3286e-05\n",
      "Epoch 147/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 519ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 2.3054e-05\n",
      "Epoch 148/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.2823e-05\n",
      "Epoch 149/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.2595e-05\n",
      "Epoch 150/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.2369e-05\n",
      "Epoch 151/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 2.2145e-05\n",
      "Epoch 152/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 522ms/step - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.1924e-05\n",
      "Epoch 153/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 533ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.1704e-05\n",
      "Epoch 154/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.1487e-05\n",
      "Epoch 155/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 534ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 2.1273e-05\n",
      "Epoch 156/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.1060e-05\n",
      "Epoch 157/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0087 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.0849e-05\n",
      "Epoch 158/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 520ms/step - loss: 0.0087 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 2.0641e-05\n",
      "Epoch 159/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 520ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 2.0434e-05\n",
      "Epoch 160/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 2.0230e-05\n",
      "Epoch 161/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 2.0028e-05\n",
      "Epoch 162/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0087 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0087 - semantic_0_loss: 0.0023 - semantic_1_loss: 0.0054 - semantic_2_loss: 0.0011 - lr: 1.9827e-05\n",
      "Epoch 163/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 1.9629e-05\n",
      "Epoch 164/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 537ms/step - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 1.9433e-05\n",
      "Epoch 165/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 1.9239e-05\n",
      "Epoch 166/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 1.9046e-05\n",
      "Epoch 167/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 518ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 1.8856e-05\n",
      "Epoch 168/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 532ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 1.8667e-05\n",
      "Epoch 169/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0083 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 21s 517ms/step - loss: 0.0083 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.8480e-05\n",
      "Epoch 170/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 524ms/step - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0011 - lr: 1.8296e-05\n",
      "Epoch 171/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 1.8113e-05\n",
      "Epoch 172/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 520ms/step - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 1.7932e-05\n",
      "Epoch 173/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.7752e-05\n",
      "Epoch 174/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 537ms/step - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0010 - lr: 1.7575e-05\n",
      "Epoch 175/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0011 - lr: 1.7399e-05\n",
      "Epoch 176/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.7225e-05\n",
      "Epoch 177/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 1.7053e-05\n",
      "Epoch 178/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 1.6882e-05\n",
      "Epoch 179/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 522ms/step - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0011 - lr: 1.6713e-05\n",
      "Epoch 180/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 521ms/step - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.6546e-05\n",
      "Epoch 181/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 543ms/step - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.6381e-05\n",
      "Epoch 182/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 536ms/step - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.6217e-05\n",
      "Epoch 183/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.6055e-05\n",
      "Epoch 184/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 527ms/step - loss: 0.0086 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0053 - semantic_2_loss: 0.0011 - lr: 1.5894e-05\n",
      "Epoch 185/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 522ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 1.5735e-05\n",
      "Epoch 186/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.5578e-05\n",
      "Epoch 187/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.5422e-05\n",
      "Epoch 188/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 531ms/step - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0010 - lr: 1.5268e-05\n",
      "Epoch 189/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0050 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 523ms/step - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0050 - semantic_2_loss: 0.0010 - lr: 1.5115e-05\n",
      "Epoch 190/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 529ms/step - loss: 0.0085 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 1.4964e-05\n",
      "Epoch 191/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 535ms/step - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.4814e-05\n",
      "Epoch 192/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.4666e-05\n",
      "Epoch 193/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 526ms/step - loss: 0.0083 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.4520e-05\n",
      "Epoch 194/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0050 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 537ms/step - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0050 - semantic_2_loss: 0.0010 - lr: 1.4374e-05\n",
      "Epoch 195/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0081 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0050 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 530ms/step - loss: 0.0081 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0050 - semantic_2_loss: 0.0010 - lr: 1.4231e-05\n",
      "Epoch 196/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0083 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 539ms/step - loss: 0.0083 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.4088e-05\n",
      "Epoch 197/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 534ms/step - loss: 0.0084 - semantic_0_loss: 0.0022 - semantic_1_loss: 0.0052 - semantic_2_loss: 0.0011 - lr: 1.3948e-05\n",
      "Epoch 198/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0050 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 528ms/step - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0050 - semantic_2_loss: 0.0010 - lr: 1.3808e-05\n",
      "Epoch 199/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 525ms/step - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.3670e-05\n",
      "Epoch 200/200\n",
      "41/41 [==============================] - ETA: 0s - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010WARNING:tensorflow:Can save best model only with val_loss available, skipping.\n",
      "41/41 [==============================] - 22s 524ms/step - loss: 0.0082 - semantic_0_loss: 0.0021 - semantic_1_loss: 0.0051 - semantic_2_loss: 0.0010 - lr: 1.3533e-05\n"
     ]
    }
   ],
   "source": [
    "# fit the model\n",
    "print('Training on', count_gpus(), 'GPUs.')\n",
    "\n",
    "train_callbacks = get_callbacks(\n",
    "    model_path,\n",
    "    lr_sched=lr_sched,\n",
    "#     monitor=\"val_loss\",\n",
    "    # monitor='loss', # training loss\n",
    "    verbose=1)          \n",
    "            \n",
    "loss_history = model.fit(\n",
    "    train_data,\n",
    "    steps_per_epoch=train_data.y.shape[0] // batch_size, # \n",
    "    epochs=epochs, \n",
    "#     validation_data=val_data,\n",
    "#     validation_steps=val_data.y.shape[0] // batch_size,\n",
    "    callbacks=train_callbacks)\n",
    "\n",
    "model.save_weights(model_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "42277d43",
   "metadata": {},
   "source": [
    "**Make prediction on K's data.**<br>\n",
    "Using NuclearSegmentation allows setting image_mpp, which has a substantial influence on performance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "dbf36df1",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T22:30:36.335826Z",
     "iopub.status.busy": "2022-11-05T22:30:36.333032Z",
     "iopub.status.idle": "2022-11-05T22:30:43.215436Z",
     "shell.execute_reply": "2022-11-05T22:30:43.216368Z"
    }
   },
   "outputs": [],
   "source": [
    "prediction_model = PanopticNet(\n",
    "    backbone='resnet50',\n",
    "    norm_method='whole_image',\n",
    "    num_semantic_classes=[1, 1], # inner distance, outer distance\n",
    "    input_shape= (512,512,1)\n",
    ")\n",
    "\n",
    "prediction_model.load_weights(model_path, by_name=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "37a766dc",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T22:30:43.225568Z",
     "iopub.status.busy": "2022-11-05T22:30:43.224427Z",
     "iopub.status.idle": "2022-11-05T22:30:43.228494Z",
     "shell.execute_reply": "2022-11-05T22:30:43.229454Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('semantic_0', (None, 512, 512, 1)), ('semantic_1', (None, 512, 512, 1))]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from deepcell.applications import NuclearSegmentation\n",
    "app = NuclearSegmentation(prediction_model)\n",
    "[(layer.name, layer.output_shape) for layer in filter(lambda x: x.name.startswith('semantic_'), app.model.layers)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "20301e2d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T22:30:43.234158Z",
     "iopub.status.busy": "2022-11-05T22:30:43.233519Z",
     "iopub.status.idle": "2022-11-05T22:30:43.244524Z",
     "shell.execute_reply": "2022-11-05T22:30:43.245224Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/home/yfong/deeplearning/dense_cell_segmentation\n",
      "['M926910_CFL_Position7_CD3_test_img.png', 'M872956_JML_Position10_CD3_test_img.png', 'M872956_JML_Position8_CD8_test_img.png', 'M926910_CFL_Position13_CD3_test_img.png', 'M872956_JML_Position8_CD3_test_img.png', 'M872956_JML_Position8_CD4_test_img.png', 'M872956_JML_Position9_CD3_test_img.png']\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "print(os.getcwd())\n",
    "INPUT_PATH=\"images/test/\"\n",
    "FILENAMES = [f for f in os.listdir(\"images/training/testimages\")]\n",
    "print(FILENAMES)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "a2eaeaa5",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T22:30:43.258476Z",
     "iopub.status.busy": "2022-11-05T22:30:43.257478Z",
     "iopub.status.idle": "2022-11-05T22:35:07.437080Z",
     "shell.execute_reply": "2022-11-05T22:35:07.438001Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "M926910_CFL_Position7_CD3_test_img.png\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/app/software/DeepCell/0.11.1-foss-2021b-CUDA-11.4.1/lib/python3.9/site-packages/deepcell_toolbox/deep_watershed.py:179: FutureWarning: `selem` is a deprecated argument name for `h_maxima`. It will be removed in version 1.0.Please use `footprint` instead.\n",
      "  markers = h_maxima(image=maxima,\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "M872956_JML_Position10_CD3_test_img.png\n",
      "M872956_JML_Position8_CD8_test_img.png\n",
      "M926910_CFL_Position13_CD3_test_img.png\n",
      "M872956_JML_Position8_CD3_test_img.png\n",
      "M872956_JML_Position8_CD4_test_img.png\n",
      "M872956_JML_Position9_CD3_test_img.png\n",
      "[[0.41237113402061853, 0.4224683544303797, 0.42410714285714285, 0.42411347517730497, 0.4038961038961039, 0.3621495327102804], [0.3973634651600753, 0.44206773618538325, 0.47332185886402756, 0.47317073170731705, 0.4245562130177515, 0.3684210526315789], [0.42857142857142855, 0.47641509433962265, 0.4772727272727273, 0.4297520661157025, 0.40625, 0.3411371237458194], [0.30666666666666664, 0.3556701030927835, 0.36342592592592593, 0.3554083885209713, 0.33739837398373984, 0.3171171171171171], [0.49245283018867925, 0.5505415162454874, 0.578397212543554, 0.533442088091354, 0.5076687116564417, 0.43626062322946174], [0.6207865168539326, 0.6015625, 0.5907990314769975, 0.5244444444444445, 0.47520661157024796, 0.4329501915708812], [0.3747276688453159, 0.4288747346072187, 0.42231075697211157, 0.45857418111753373, 0.41563055062166965, 0.3890728476821192]]\n",
      "[1.  1.2 1.4 1.6 1.8 2. ]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[3, 2, 2, 2, 2, 0, 3]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# more detailed mapping of image_mpp parameter\n",
    "mpps=np.linspace(start=1, stop=2, num=6)\n",
    "\n",
    "AP_mat=[]\n",
    "for CURR_IM_NAME in FILENAMES:\n",
    "    print(CURR_IM_NAME)\n",
    "    im0 = io.imread(os.path.join(INPUT_PATH, CURR_IM_NAME))\n",
    "    mask_true=io.imread(os.path.join(INPUT_PATH, CURR_IM_NAME.replace(\"img\",\"masks\")))\n",
    "\n",
    "    x = np.expand_dims(im0, axis=-1)\n",
    "    x = np.expand_dims(x, axis=0)\n",
    "    y, tile_info = app._tile_input(x)\n",
    "    AP_arr=[]\n",
    "    for mpp in mpps:\n",
    "        prd = app._untile_output(app.predict(y, image_mpp=mpp), tile_info)[0,:,:,0]\n",
    "        a=syotil.csi(mask_true, prd)\n",
    "        AP_arr.append(a)\n",
    "    AP_mat.append(AP_arr)\n",
    "print(AP_mat)\n",
    "print(mpps)\n",
    "[x.index(np.max(x)) for x in AP_mat]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "5a5a611e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T22:35:07.447441Z",
     "iopub.status.busy": "2022-11-05T22:35:07.446296Z",
     "iopub.status.idle": "2022-11-05T22:35:23.050102Z",
     "shell.execute_reply": "2022-11-05T22:35:23.051028Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "{'M926910_CFL_Position7_CD3_test_img.png': 0.41237113402061853, 'M872956_JML_Position10_CD3_test_img.png': 0.3973634651600753, 'M872956_JML_Position8_CD8_test_img.png': 0.42857142857142855, 'M926910_CFL_Position13_CD3_test_img.png': 0.30666666666666664, 'M872956_JML_Position8_CD3_test_img.png': 0.49245283018867925, 'M872956_JML_Position8_CD4_test_img.png': 0.6207865168539326, 'M872956_JML_Position9_CD3_test_img.png': 0.3747276688453159, 'mAP': 0.4332771014723882}\n"
     ]
    }
   ],
   "source": [
    "mpp_select=1.0\n",
    "\n",
    "APs={}\n",
    "for CURR_IM_NAME in FILENAMES:\n",
    "    im0 = io.imread(os.path.join(INPUT_PATH, CURR_IM_NAME))\n",
    "    mask_true=io.imread(os.path.join(INPUT_PATH, CURR_IM_NAME.replace(\"img\",\"masks\")))\n",
    "\n",
    "    x = np.expand_dims(im0, axis=-1)\n",
    "    x = np.expand_dims(x, axis=0)\n",
    "    y, tile_info = app._tile_input(x)\n",
    "    print(x.shape)\n",
    "    print(y.shape)\n",
    "    print(tile_info)\n",
    "    pred = app.predict(y, image_mpp=mpp_select)\n",
    "    prd = app._untile_output(pred, tile_info)\n",
    "    #io.imshow(prd[0,:,:,0])\n",
    "    plt.show()\n",
    "    \n",
    "    APs[CURR_IM_NAME] = syotil.csi(mask_true, prd[0,:,:,0])# masks may lose one pixel if dimension is odd pixels\n",
    "\n",
    "APs[\"mAP\"]=np.mean(list(APs.values()))\n",
    "print(APs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "f31fa740",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T22:35:23.059399Z",
     "iopub.status.busy": "2022-11-05T22:35:23.058288Z",
     "iopub.status.idle": "2022-11-05T22:35:52.485498Z",
     "shell.execute_reply": "2022-11-05T22:35:52.486422Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "(1, 1040, 233, 1)\n",
      "(1, 1568, 512, 1)\n",
      "{'padding': True, 'x_pad': (264, 264), 'y_pad': (139, 140)}\n",
      "{'M926910_CFL_Position7_CD3_test_img.png': 0.42725880551301687, 'M872956_JML_Position10_CD3_test_img.png': 0.4614035087719298, 'M872956_JML_Position8_CD8_test_img.png': 0.5124378109452736, 'M926910_CFL_Position13_CD3_test_img.png': 0.3349282296650718, 'M872956_JML_Position8_CD3_test_img.png': 0.5732368896925859, 'M872956_JML_Position8_CD4_test_img.png': 0.6192893401015228, 'M872956_JML_Position9_CD3_test_img.png': 0.4495798319327731, 'mAP': 0.48259063094602483}\n"
     ]
    }
   ],
   "source": [
    "mpp_select=1.3\n",
    "\n",
    "APs={}\n",
    "for CURR_IM_NAME in FILENAMES:\n",
    "    im0 = io.imread(os.path.join(INPUT_PATH, CURR_IM_NAME))\n",
    "    mask_true=io.imread(os.path.join(INPUT_PATH, CURR_IM_NAME.replace(\"img\",\"masks\")))\n",
    "\n",
    "    x = np.expand_dims(im0, axis=-1)\n",
    "    x = np.expand_dims(x, axis=0)\n",
    "    y, tile_info = app._tile_input(x)\n",
    "    print(x.shape)\n",
    "    print(y.shape)\n",
    "    print(tile_info)\n",
    "    pred = app.predict(y, image_mpp=mpp_select)\n",
    "    prd = app._untile_output(pred, tile_info)\n",
    "    #io.imshow(prd[0,:,:,0])\n",
    "    plt.show()\n",
    "    \n",
    "    APs[CURR_IM_NAME] = syotil.csi(mask_true, prd[0,:,:,0])# masks may lose one pixel if dimension is odd pixels\n",
    "\n",
    "APs[\"mAP\"]=np.mean(list(APs.values()))\n",
    "print(APs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "7bc06419",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-11-05T22:35:52.495325Z",
     "iopub.status.busy": "2022-11-05T22:35:52.494226Z",
     "iopub.status.idle": "2022-11-05T22:35:52.531968Z",
     "shell.execute_reply": "2022-11-05T22:35:52.531199Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                         0         1\n",
      "0   M926910_CFL_Position7_CD3_test_img.png  0.427259\n",
      "1  M872956_JML_Position10_CD3_test_img.png  0.461404\n",
      "2   M872956_JML_Position8_CD8_test_img.png  0.512438\n",
      "3  M926910_CFL_Position13_CD3_test_img.png  0.334928\n",
      "4   M872956_JML_Position8_CD3_test_img.png  0.573237\n",
      "5   M872956_JML_Position8_CD4_test_img.png  0.619289\n",
      "6   M872956_JML_Position9_CD3_test_img.png   0.44958\n",
      "7                                      mAP  0.482591\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "df = pd.DataFrame([FILENAMES+[\"mAP\"], list(APs.values())])\n",
    "print(df.transpose())\n",
    "df.to_csv('images/training/csi_'+model_name+'.txt', index=False, header=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "520fe391",
   "metadata": {},
   "source": [
    "**mAP**<br>\n",
    "image_mpp=1:   0.43<br>\n",
    "image_mpp=1.3: 0.48<br>"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}