v0.2.0

Release Notes for Version 0.2.0

This version has all my set of experiments for the breast density classification in BI-RADS®.. The primary objective of these experiments is to evaluate the model's performance in distinguishing between the density classes defined in BI-RADS®.

Key Features

The initial experiments have been conducted, and the results are now accessible for review and analysis. You can find all the experiment data and results in the root folder under experiments/

Model Weights Availability:

The best-performing model weights from these binary classification experiments are now available for download. These weights have been uploaded to Kaggle for ease of access and further use in your projects, and can be downloaded here, for the binary model, and here for the multiclass models.

Known bugs

In this version and the others before, some models were saved in torch.nnDataParallel mode, so it is possible that, when loading the model, the weights dictionary contains a prefix for each layer. In order to fix this, load the model with DataParallel or uncomment some of the lines in function test() at src/workflow/binary/model_tester.py or at src/workflow/multiclass/model_tester.py

v0.1.1

Release Notes for Version 0.1.1

This version has all my set of experiments focused on the binary model approach for the breast density classification in BI-RADS®.. The primary objective of these experiments is to evaluate the model's performance in distinguishing between the density classes defined in BI-RADS®.

Key Features

The initial experiments have been conducted, and the results are now accessible for review and analysis. You can find all the experiment data and results in the root folder under experiments/binary.

Model Weights Availability:

The best-performing model weights from these binary classification experiments are now available for download. These weights have been uploaded to Kaggle for ease of access and further use in your projects, and can be downloaded here.

Known bugs

In this version and the others before, some models were saved in torch.nnDataParallel mode, so it is possible that, when loading the model, the weights dictionary contains a prefix for each layer. In order to fix this, load the model with DataParallel or uncomment some of the lines in function test() at src/workflow/binary/model_tester.py.

Future Work:

The multiclass classification training algorithm will be available at version 0.2.0.

v0.1.0

Release Notes for Version 0.1.0

This version introduces my first set of experiments focused on binary classification. The primary objective of these experiments is to evaluate the model's performance in distinguishing between the density classes defined in BI-RADS®.

Key Features

The initial experiments have been conducted, and the results are now accessible for review and analysis. You can find all the experiment data and results in the root folder under experiments/binary.

Model Weights Availability:

The best-performing model weights from these binary classification experiments are now available for download. These weights have been uploaded to Kaggle for ease of access and further use in your projects, and can be downloaded here.

Future Work:

The multiclass classification training algorithm will be available at version 0.2.0.