new release

.gitignore

@@ -13,7 +13,7 @@ __pycache__
 **/.ipynb_checkpoints/
 
 # Ignoring the default events directory.
-*/outputs
+keyword_information_extraction/88_33_recall_outputs
 
 # Editor temporaries
 *.swn

README.md

@@ -1,7 +1,7 @@
 # ICDAR 2019 Robust Reading Challenge on Scanned Receipts OCR and Information Extraction
 
 > I submitted my results to the competition with the last model checkpoint.
-> Unfortunately, my results are not reproducibles accross different machines but were the same on my machine through multiple executions.
+> Unfortunately, my results are not reproducibles across different machines but were the same on my machine through multiple executions.
 > Indeed, after having seeded the training, the results obtained were slightly different on the borrowed NVIDIA GTX 1080.
 > This is not surprising because it is mentioned in the Pytorch documentation about [reproducibility](https://pytorch.org/docs/stable/notes/randomness.html).
 > That is why I gave up on seeding the training.
@@ -96,22 +96,26 @@ in `task1_2_test(361p)`. Indeed, the number of files in `task1_2_test(361p)` and
 same (360 and 361 respectively). The reason is that this filename `X51006619570.jpg` is missing, and it turns out that
 it was in `task3-test 347p) -`.
 
-Another mistake lies within the folder `0325updated.task2train(626p)` which is used for
-the [Task 3: Keyword Information Extraction](./keyword_information_extraction). Indeed, there are three files for which
-the date format is wrong and here are the corrections that were made:
+Another mistakes lie within the folder `0325updated.task2train(626p)` and `0325updated.task1train(626p)` which are used for
+the [Task 3: Keyword Information Extraction](./keyword_information_extraction). Indeed, there are files for which
+the date format, address and company are wrong and here are the corrections that were made:
 
-|   Filename   |          Correction           |
-| :----------: | :-----------------------------|
-| X51005447850 | Turn 20180304 into 04/03/2018 |
-| X51005715010 | Turn 25032018 into 25/03/2018 |
-| X51006466055 | Turn 20180428 into 28/04/2018 |
+|                          Directory                          |   Filename   | Correction                                                                         |
+|:-----------------------------------------------------------:|:------------:|:-----------------------------------------------------------------------------------|
+|                 0325updated.task2train(626p                 | X51005447850 | Change 20180304 into 04/03/2018                                                    |
+| 0325updated.task1train(626p) & 0325updated.task2train(626p) | X51005715010 | Change 25032018 into 25.03.2018                                                    |
+| 0325updated.task1train(626p) & 0325updated.task2train(626p) | X51006466055 | Change 20180428 into 2018-04-28                                                    |
+|                0325updated.task2train(626p)                 | X51008114284 | Remove one occurrence of 'KAWASAN PERINDUSTRIAN BALAKONG,' from the address        |
+|                0325updated.task2train(626p)                 | X00016469620 | Remove " (MR DIY TESCO TERBAU)" from the address                                   |
+|                0325updated.task2train(626p)                 | X00016469623 | Remove " (TESCO PUTRA NILAI)" from the address                                     |
+|                0325updated.task2train(626p)                 | X51006502531 | Change "FAMILYMART" into "MAXINCOME RESOURCES SDN BHD (383322-D)" from the company |
 
 #### Downloads <a name="dset-downloads"/>
 
 The **Original dataset** can be found [Google Drive](https://drive.google.com/open?id=1ShItNWXyiY1tFDM5W02bceHuJjyeeJl2)
 or [Baidu NetDisk](https://pan.baidu.com/s/1a57eKCSq8SV8Njz8-jO4Ww#list/path=%2FSROIE2019&parentPath=%2F).
 
-Taking into account to what was mentionned in [dataset mistakes](#dset-mistakes) above, you may obviously want to make
+Taking into account to what was mentioned in [dataset mistakes](#dset-mistakes) above, you may obviously want to make
 the changes by your own, but I have already made the corrections, and it can be downloaded via the bash
 script: [sroie2019.sh](scripts/datasets/sroie2019.sh) and here is how to run it:
 
@@ -129,7 +133,7 @@ script: [sroie2019.sh](scripts/datasets/sroie2019.sh) and here is how to run it:
 
   Do not forget to specify the new directory inside this [file](scripts/datasets/dataset_roots.py).
 
-For Windows users who do not have bash on their system, you may want to
+For Windows users who do not have `bash` on their system, you may want to
 install [git bash](https://git-scm.com/download/win). Once it is installed, you can set the entire `git bin` folder in
 the [environment variables](https://stackoverflow.com/questions/17312348/how-do-i-set-windows-environment-variables-permanently)
 .
@@ -140,17 +144,17 @@ Here are methods used for the competition. Inside each folder representing the t
 the proposed method and the training, demo and evaluation procedures as well.
 
 - **Task 1 - Text Localization**: Connectionist Text Proposal Network (CTPN).
-- **Task 3 - Keyword Information Extraction**: Character-Aware CNN + Highway + BiLSTM (Char LM).
+- **Task 3 - Keyword Information Extraction**: Character-Aware CNN + Highway + BiLSTM (CharLM).
 
 ## Results <a name="results"/>
 
 The results are listed as follows (Note that for the task 3, I manually fix each and every OCR mismatches for fair
 comparison results):
 
-|  Task  |   Recall  | Precision |  Hmean   | Evaluation Method  |     Model      |   Parameters   | Model Size | Weights |
-| :----: | :-------: | :-------: | :------: | :----------------: | :------------: | :------------: | :--------: | :--------------------------------------------------------------------------------------------------------------------------------- |
-| Task 1 |   97.16%  |   97.10%  |  97.13%  |      Deteval       |     CTPN       |   16,900,032   |  268.3 MB  | [Last checkpoint](https://github.com/BlackStar1313/ICDAR-2019-RRC-SROIE/releases/download/v1.0/CTPN_FINAL_CHECKPOINT.pth) |
-| task 3 |   96.18%  |   97.45%  |  96.81%  |         /          |     Char LM    |   4,740,590    |  75.9 MB   | [last checkpoint](https://github.com/BlackStar1313/ICDAR-2019-RRC-SROIE/releases/download/v1.0/CHARLM_CNN_HIGHWAY_LSTM_FINAL_CHECKPOINT.pth) |
+|  Task  | Recall | Precision | Hmean  | Evaluation Method | Model  | Parameters | Model Size | Weights                                                                                                                                      |
+|:------:|:------:|:---------:|:------:|:-----------------:|:------:|:----------:|:----------:|:---------------------------------------------------------------------------------------------------------------------------------------------|
+| Task 1 | 97.52% |  97.40%   | 97.46% |      Deteval      |  CTPN  | 18,450,332 |  268.3 MB  | [Last checkpoint](https://github.com/BlackStar1313/ICDAR-2019-RRC-SROIE/releases/download/v1.0/CTPN_FINAL_CHECKPOINT.pth)                    |
+| task 3 | 98.20% |  98.48%   | 98.34% |         /         | CharLM | 4,740,590  |  75.9 MB   | [last checkpoint](https://github.com/BlackStar1313/ICDAR-2019-RRC-SROIE/releases/download/v1.0/CHARLM_CNN_HIGHWAY_LSTM_FINAL_CHECKPOINT.pth) |
 
 ## User guide <a name="user-guide"/>
 
@@ -166,6 +170,12 @@ Besides, all the experiments and results were performed on my personal gaming co
 - NVIDIA® GeForce® RTX 2070 SUPER™, 8Go GDDR6
 - OS: Dual boot Windows/Ubuntu 20.04
 
+and DIVA GPU cluster:
+
+- 9th Gen Intel(R) Core(TM) i9 9900K (8-core, 16 MB cache)
+- NVIDIA® GeForce® GTX 1080
+- OS: Ubuntu 18.04
+
 ### Conda environment setup <a name="environment-setup"/>
 
 For Mac, Windows and Linux users, if `conda` is not installed, then you need to follow
@@ -187,10 +197,10 @@ this [documentation](https://docs.continuum.io/anaconda/install/).
 
    This will create a new conda environment named `SROIE2019` on your system, which will give you all the packages
    needed for this repo. If you do not own any NVIDIA GPUs (with CUDA capable-system), then you must remove
-   the `cudatoolkit` and `cudnn` lines in the [environment.yml](env/environment.yml) file. Otherwise, make sure your graphic card
-   supports the installed version of CUDA.
+   the `cudatoolkit` and `cudnn` lines in the [environment.yml](env/environment.yml) file. 
+   Otherwise, make sure your graphic card supports the installed version of CUDA.
 
-4. Activating the new environment
+3. Activating the new environment
 
    ```
    conda activate SROIE2019
@@ -204,7 +214,7 @@ this [documentation](https://docs.continuum.io/anaconda/install/).
 
    If you want to deactivate the environment, you can simply do: `conda deactivate`
 
-5. Verify that the new environment was installed correctly
+4. Verify that the new environment was installed correctly
 
    ```
    conda env list
@@ -222,7 +232,7 @@ the [manager](https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/ma
 
 ### Visualizer
 
-To use vidsom, you must make sure the server is running before you run the training.
+To use `Vidsom`, you must make sure the server is running before you run the training.
 
 #### Starting the server with
 

env/environment.yml

@@ -1,23 +1,20 @@
 name: SROIE2019
 channels:
-  - conda-forge
+  - pytorch
+  - defaults
 dependencies:
-  - python=3.7
   - pip
-  - numpy
+  - cudnn
   - pytorch
+  - python=3.7
   - torchvision
   - cudatoolkit
-  - cudnn
-  - opencv
-  - python-Levenshtein
-  - scikit-learn
-  - fuzzywuzzy
-  - pyyaml
-  - visdom
-  - jsonpatch
-  - pytorch-model-summary
-  - regex
-  - yacs
   - pip:
-      - vizer
+      - yacs
+      - regex
+      - numpy
+      - visdom
+      - PyYAML
+      - tabulate
+      - opencv-python
+      - pytorch-model-summary

functional/data/dataloader/__init__.py

@@ -0,0 +1 @@
+from .data_loader import dataloader as Dataloader

functional/data/dataloader/data_loader.py

@@ -0,0 +1,43 @@
+import torch
+
+from typing import Optional
+from torch.utils.data import Dataset, DataLoader, RandomSampler, SequentialSampler, BatchSampler
+
+
+def dataloader(dataset: Dataset, is_train: Optional[bool] = False, **kwargs):
+    """
+    Create the :class:`torch.utils.data.DataLoader` with given parameters.
+
+    Args:
+        dataset (Dataset): The dataset to iterate through.
+        is_train (bool , optional): A boolean to check whether the training mode is activated or not.
+        **kwargs: The dataloader parameters.
+        
+    Returns:
+        An iterable over the given dataset.
+        
+    """
+    shuffle = kwargs.get("shuffle", None)
+
+    if not is_train or shuffle is None or not shuffle:
+        sampler = SequentialSampler(data_source=dataset)
+    else:
+        generator = kwargs.get("generator", None)
+        if generator is None:
+            generator = torch.Generator(device=torch.device("cpu"))
+        sampler = RandomSampler(dataset, generator=generator)
+
+    batch_size = kwargs["batch_size"]
+    drop_last = kwargs["drop_last"]
+    batch_sampler = BatchSampler(sampler=sampler, batch_size=batch_size, drop_last=drop_last)
+
+    # Once we set the following arguments below, we must remove them. Otherwise, we will get a ValueError:
+    # batch_sampler option is mutually exclusive with batch_size, shuffle, sampler, and drop_last
+    kwargs.pop("shuffle")
+    kwargs.pop("batch_size")
+    kwargs.pop("drop_last")
+
+    kwargs["batch_sampler"] = batch_sampler
+    dataloader = DataLoader(dataset=dataset, **kwargs)
+
+    return dataloader

functional/data/preprocessing/__init__.py

@@ -0,0 +1 @@
+from .split_labels import crop_preprocessing