Major release with the paper

What's Changed

• Implemented a common copyright message for CLI and updated links & email by @sarthakpati in #512
• Change python-test.yml to pass on docs-only by @AlexanderGetka-cbica in #517
• Making the flowchart a bit more generic to showcase inference by @sarthakpati in #514
• Updated install verification logic by @sarthakpati in #520
• Updated implementation for augmentations that need std-dev by @sarthakpati in #521
• Ability to generate multiple training configurations using a specific strategy by @sarthakpati in #525
• Updated patch divisibility logic by @sarthakpati in #523
• Added sanity check for norm_type for non-vgg network architectures by @sarthakpati in #528
• Fixing requirement for SimpleITK for a problematic version by @sarthakpati in #537
• Preprocessing is now working when a random set of metadata columns are present by @sarthakpati in #535
• Added ability to combine classification inference results from different architectures by @sarthakpati in #531
• Added ability to save initial and latest models in addition to best by @sarthakpati in #539
• Latest model is now saved correctly by @sarthakpati in #546
• Consideration of scaling_factor has been moved to loss/metrics by @sarthakpati in #541
• Added ability to specify testing data csv in main cli by @sarthakpati in #545
• added more information for #529 by @sarthakpati in #549
• Add docker docs by @AlexanderGetka-cbica in #550
• Added more details to documentation specifically related to training customization by @sarthakpati in #552
• Added more details about Docker by @sarthakpati in #554
• Separate modelDir and outputDir parameters for inference to cohesiveness by @sarthakpati in #557
• The logs now include metric headers from overall_stats by @sarthakpati in #558
• Fix parseConfig separately from my other PR by @AlexanderGetka-cbica in #561

Full Changelog: 0.0.5...0.0.16

Major release before dependency updates.

What's Changed

• Python 3.8 fixes by @AlexanderGetka-cbica in #453
• Version updated for development by @sarthakpati in #452
• Fixing the stride_size check for histology inference by @sarthakpati in #455
• Update CI (PyTest to 3.8, and Docker Image creation) by @AlexanderGetka-cbica in #461
• Bump numpy from 1.21.0 to 1.22.0 by @dependabot in #460
• Add collectStats functionality for classification tasks by @garvm7 in #463
• Ensuring params are populated correctly even if only testing data is invoked by @sarthakpati in #466
• Added blending probability maps with original WSI by @sarthakpati in #465
• Update data type for stride size by @sarthakpati in #458
• Add FP16 option for model optimization by @junwenwu in #457
• Added option to not pick up pretrained weights for imagenet_vgg by @sarthakpati in #469
• Added ability to print model size and summary at run-time by @sarthakpati in #468
• Better handling of large WSIs by @sarthakpati in #472
• Added exception testing for histology classification by @sarthakpati in #477
• Fixing reported Codacy fixes by @sarthakpati in #474
• Formatting fixes for readability by @sarthakpati in #478
• Completely excising VIPS from GaNDLF by @sarthakpati in #480
• Update test workflow with file path filtering by @AlexanderGetka-cbica in #473
• Saving report for failure cases from pytest as a separate log by @sarthakpati in #482
• New metric added to calculated per-class accuracy by @sarthakpati in #486
• Single utility function to return array from tensor/image by @sarthakpati in #483
• Increasing unit test coverage by @sarthakpati in #487
• Fixed the order of coordinates in predictions.csv for WSI inference by @sarthakpati in #490
• Added ability to call specific post processing algorithms after reverse one-hot encoding by @sarthakpati in #494
• Adding Largest connected component analysis as a post-processing step by @Geeks-Sid in #467
• Dedicated rescaling preprocessing function added for increased flexibility by @sarthakpati in #496
• Fixes multi-batch issue for per_label_accuracy by @sarthakpati in #498
• Included classification-specific metrics using overall predictions and ground truths by @sarthakpati in #500
• version update for tagging by @sarthakpati in #501

New Contributors

• @garvm7 made their first contribution in #463

Full Changelog: 0.0.14...0.0.15

Major release with lots of new functionality

What's Changed

• updated version for development by @sarthakpati in #259
• Fix for get_final_layer use in GANDLF.utils by @sarthakpati in #260
• Add functionality to save training patches by @sarthakpati in #263
• Updates to integrate new OPM APIs by @sarthakpati in #261
• Added option to print per-label segmentation scores by @sarthakpati in #264
• updated initialization for dicts by @sarthakpati in #267
• Motion artifact separated by @sarthakpati in #270
• added black and opencv-python as an explicit requirements by @sarthakpati in #271
• Bump numpy from 1.19.2 to 1.21.0 by @dependabot in #272
• Added capability to use InstanceNorm in DenseNet by @sarthakpati in #274
• Updated implementations of VGG and DenseNet to use ModelBase for consistency by @sarthakpati in #276
• Added git hash to model save by @sarthakpati in #277
• Added FAQ by @sarthakpati in #279
• Update to how git_hash is saved by @sarthakpati in #280
• Documentation updated for new module structure by @sarthakpati in #283
• Accuracy is now standardized from torchmetrics by @sarthakpati in #287
• Adding automated linting check for black by @sarthakpati in #284
• Post-processing module added by @sarthakpati in #289
• Added a GANDLF.anonymize module by @sarthakpati in #291
• Updated documentation (faq and usage) by @sarthakpati in #294
• Git submodule update is now integrate in setup by @sarthakpati in #296
• Ensure colorjitter works with defaults by @sarthakpati in #297
• More progress bars for better feedback by @sarthakpati in #301
• Update usage.md - plot final results by @anishshriram in #305
• Using Resample and Resize from TorchIO by @Geeks-Sid in #293
• Ensure hole filling output is int by @sarthakpati in #306
• Added instructions for dependency management by @sarthakpati in #307
• Minor fix for reset_prev by @sarthakpati in #310
• added captk and fets links by @sarthakpati in #311
• Updates documentation based on SB's notes by @sarthakpati in #313
• Added script to verify installation by @sarthakpati in #315
• Added ability to save preprocessed files with data augmentation by @sarthakpati in #317
• Added ResNet with multiple configurations by @Sofia-Mouchtaris in #319
• Bug fix in parameters for gandlf_collectStats by @dlang95 in #321
• Resample_Min added to ensure a baseline resolution for all images getting processed by @sarthakpati in #325
• Added final predictions for classification by @sarthakpati in #328
• Ensuring cropping zero planes are applied for preprocessing by @sarthakpati in #329
• Typo fixed for hd95 by @sarthakpati in #331
• Using TiffSlide instead of OpenSlide for a better deployment by @sarthakpati in #303
• Histology anonymization added by @Geeks-Sid in #281
• Bug fix for gandlf_collectStats by @sarthakpati in #335
• Adding resize_image as a separate option as compared to resize_patch by @sarthakpati in #337
• Sanity checking is now softer with a threshold by @sarthakpati in #340
• Added ability to resume with and without updating parameters and/or data by @sarthakpati in #344
• Reverse one-hot logic is now updated to output unique labels by @sarthakpati in #342
• Added EfficientNet with multiple configurations by @Sofia-Mouchtaris in #327
• Links added for current networks by @sarthakpati in #346
• Add Dockerfiles and Docker Hub CI by @AlexanderGetka-cbica in #353
• Add GitHub Packages upload step, refine Docker CI further by @AlexanderGetka-cbica in #355
• Added Docker Image builds by @sarthakpati in #354
• docker ci badge position change by @sarthakpati in #356
• Docker CI: Fix login requirement -- no need to log in for PRs (oops!) by @AlexanderGetka-cbica in #359
• Added some checks for inference mode by @sarthakpati in #358
• Fix for invalid classification accuracy in training by @orhunguley in #361
• Post-training optimization using OpenVINO added by @junwenwu in #312
• Enabling inference without having access to the ground truth by @sarthakpati in #363
• Ensure ONNX-related print only happens once during training by @sarthakpati in #364
• Fixed issue with balanced_accuracy parsing by @sarthakpati in #366
• Hard requirement of PyTorch to LTS by @sarthakpati in #369
• Added post-processing step to map predicted labels to new values by @sarthakpati in #367
• Fix Docker CI on PRs again (meta information missing causing a rebuild -> out of space) by @AlexanderGetka-cbica in #371
• Fixing docker builds by @sarthakpati in #372
• ROCm Docker builds are no longer valid since we are only supporting PyTorch LTS by @sarthakpati in #370
• Custom inference loop added for histo classification by @sarthakpati in #374
• UNet and Light UNet with varying depth by @Sofia-Mouchtaris in #357
• Create codacy-coverage-reporter.yml by @sarthakpati in #368
• added a comment about padding options by @sarthakpati in #376
• Improved penalty calculation and added ImageNet-specific VGG by @Geeks-Sid in #365
• Error checks added to ImageNet VGG by @sarthakpati in #377
• updated config comment for num_channels by @sarthakpati in #378
• Logic fixed for step scheduler by @sarthakpati in #379
• Add option to save model at every epoch by @sarthakpati in #381
• Added note for classification/regression tasks by @sarthakpati in #385
• Added RGBA to RGB conversion in preprocessing by @sarthakpati in #384
• Improved default step_size calculation by @sarthakpati in #386
• Added a key for exp scheduler by @sarthakpati in #387
• Documentation updates by @sarthakpati in #388
• Fixing tests on named venv by @sarthakpati in #382
• Fixing pytorch version dependency for macOS by @sarthakpati in #390
• Add option to strategize the final aggregation for segmentation by @sarthakpati in #392
• Fixing per-label metrics display by @sarthakpati in #393
• Added submodule to hand template-matching based normalization by @sarthakpati in #394
• Added RGB to RGBA conversion for preprocessing by @sarthakpati in #397
• A new transformer-based segmentation arch (UNetR) has been added by @Sofia-Mouchtaris in #398
• Fixing different device issue on multi-layer UNet and UNETR by @sarthakpati in #399
• Added sigmoid multiplier for UNet model output by @sarthakpati in #400
• Documentation update for installation by @sarthakpati in #403
• More inference mode options added by @sarthakpati in #405
• Update one_hot and reverse_one_hot logic for succinctness by @sarthakpati in #404
• moving get_final_layer method functionality to GANDLF.utils to avoid by @brandon-edwards in #407
• API updates to make GaNDLF's compute objects callable outside the tool by @sarthakpati in #401
• Added memory usage for cuda u...

Documentation updates and minor bugfixes

From History:

• Deep supervision added
• Documentation updated
• Model IO is now standardized

Multiple new features for training

From History:

• New Network: Light-UNet
• Misc bugfixes
• Automatic check-pointing of the model has been added
• Extending the codebase has been simplified
• New optimizers added
• New metrics added
• Affine augmentation can now be significantly fine-tuned
• Update logic for penalty calculation
• RGB-specific augmentation added
• Cropping added

New metric calculation mechanism

Changelog:

• Misc bugfixes for segmentation and classification
• DFU 2021 parameter file added
• Added SDNet for supervised learning - https://doi.org/10.1016/j.media.2019.101535
• Added option to re-orient all images to canonical
• Preprocessing and augmentation made into separate submodules

Updated feature set for segmentation and classification

Changelog:

• Half-time epoch loss and metric output added for increased information
• Gradient clipping added
• Per-epoch details in validation output added
• Different types of normalization layer options added
• Hausdorff as a validation metric has been added
• New option to save preprocessed data before the training starts

Major refactor finished

• First level of code re-organization
• No API changes for config files

Final tag before major refactor

0.0.8

Merge pull request #39 from sarthakpati/master