misc.py

from __future__ import unicode_literals
from __future__ import print_function
from __future__ import division
from __future__ import absolute_import
import SimpleITK as sitk
import os
import numpy as np
import pandas as pd
import scipy.ndimage
import time
import os
import cv2
import tensorflow as tf

# TensorFlow Device Configuration 
def setup_device(gpuid=None):
    devices = []

    if gpuid is not None and not isinstance(gpuid, str): gpuid = str(gpuid)
    if gpuid is not None: nb_devices = len(gpuid.split(','))
    else:                 nb_devices = 1

    if gpuid is not None and (gpuid!='-1'):
        # Define GPU Devices for Mirrored Strategy
        if (nb_devices>1):
            for i in range(nb_devices): devices.append('/gpu:'+str(i))
        else:                           devices = '/gpu:0'
        
        # Set Number of Visible GPUs
        os.environ['CUDA_VISIBLE_DEVICES'] = gpuid

        # GPU Memory Configuration Differs Between TF 1 and 2
        if hasattr(tf, 'ConfigProto'):
            config                          = tf.ConfigProto()
            config.gpu_options.allow_growth = True
            config.allow_soft_placement     = True
            config.log_device_placement     = True
            tf.keras.backend.set_session(tf.Session(config=config))
        else:
            tf.config.set_soft_device_placement(True)
            for pd in tf.config.experimental.list_physical_devices('GPU'):
                tf.config.experimental.set_memory_growth(pd, True)
    else:
        devices = '/cpu:0'
        os.environ['CUDA_VISIBLE_DEVICES'] = '-1'

    return devices, nb_devices

# Print Overview Of Run using CL Arguments
def print_overview(args):
    # Training Objective
    if   (args.TRAIN_OBJ=='zonal'):  print("Training Objective: Anatomical (WG,TZ,PZ) Segmentation")
    elif (args.TRAIN_OBJ=='lesion'): print("Training Objective: Diagnostic (csPCa) Segmentation")
    
    # Probabilistic Outputs 
    if args.UNET_PROBABILISTIC:     print("Probabilistic Outputs: Enabled")
    else:                           print("Probabilistic Outputs: Disabled")

    # Deep Supervision
    if args.UNET_DEEP_SUPERVISION:  print("Deep Supervision: Enabled")
    else:                           print("Deep Supervision: Disabled")

    # Number of Classes at Train-Time
    if   (args.TRAIN_OBJ=='zonal'):    print("Number of Classes at Train-Time: 3")       
    elif (args.TRAIN_OBJ=='lesion'):   print("Number of Classes at Train-Time: 2")      
    
    # Training Hyperparameters
    print("Batch Size:",            args.BATCH_SIZE)
    print("Initial Learning Rate:", args.BASE_LR)
    
    # Optimizer
    if   (args.OPTIMIZER=='adam'):     print("Optimizer: Adam w/ AMSGrad")
    elif (args.OPTIMIZER=='momentum'): print("Optimizer: Stochastic Gradient Descent w/ Nesterov Momentum")
    
    # Loss Function
    if   (args.LOSS_MODE=='distribution_focal'):  print("Loss Function: Distribution/Focal Loss (alpha="+str(args.FOCAL_LOSS_ALPHA)+"; gamma="+str(args.FOCAL_LOSS_GAMMA)+")")
    elif (args.LOSS_MODE=='region_boundary'):     print("Loss Function: Soft Dice + Boundary/Surface Loss")
    
    # Train-Time Augmentations
    if (args.AUGM_PARAMS[0]>0): 
        print("Train-Time Augmentations: \
          (Master Probability: "+        str(args.AUGM_PARAMS[0])+\
        "; Transformation Probability: "+str(args.AUGM_PARAMS[1])+\
        "; X-Y Translation: "+           str(args.AUGM_PARAMS[2])+\
        "; Rotation Degree: ±"+          str(args.AUGM_PARAMS[3])+\
        "; Horizontal Flip: "+           str(args.AUGM_PARAMS[4])+\
        "; Zoom Factor: "+               str(args.AUGM_PARAMS[5])+\
        "; Gaussian Noise: stddev<="+    str(args.AUGM_PARAMS[6])+\
        "; Channel Shift: "+             str(args.AUGM_PARAMS[7])+\
        "; Simulate Poor Scan Quality: "+str(args.AUGM_PARAMS[8])+\
        "; Gamma Shifts: "+              str(args.AUGM_PARAMS[9])+")")
    else:                       print("Train-Time Augmentations: Disabled")
    
    # Dropout Mode
    if   (args.UNET_DROPOUT_MODE=='standard'):     print("Dropout: Standard (rate={:.2f})".format(args.UNET_DROPOUT_RATE))
    elif (args.UNET_DROPOUT_MODE=='monte-carlo'):  print("Dropout: Monte Carlo (rate={:.2f})".format(args.UNET_DROPOUT_RATE))
    elif (args.UNET_DROPOUT_RATE==0):              print("Dropout: Disabled")