This is the code for training the model for Module 3 of this paper.
The paper was admitted to EMBC 2020 and was published on IEEE Xplore.
The pipeline of the system from the paper is below. Module 3 is Abnormality classification. The visualization was done using Grad-Cam.
pip install -r requirements.txt
Download images and put all of them in the overall directory.
To run the training with default options, simply run python train.py
--gpus: IDs of GPUs to train on.--batch_size: Number of samples that will be propagated through the network in one forward/backward pass.--epochs: Number of epochs to train the model.--path_to_images: Directory where images are stored.--lr: Learning rate.--log_dir: Directory where logs and wieghts will be saved.--num_worker: Positive integer will turn on multi-process data loading with the specified number of loader worker processes (Check PyTorch docs).--train_list: Path to a file with image names and labels of the train data.--val_list: Path to a file with image names and labels of the validation data.
After in the training logs directory will be created. The model weights can be found in logs/experiment_name/checkpoints.
Download our model weights file to run inference (Google Drive) and place it into checkpoints.
To run the inference on the test data with default options and our weights, simply run python test.py
You can check your own model by setting a path to your .pth file by -- checkpoint.
The visualisation of the metrics in the pdf format will be saved in a directory defined by --test_outdir.
--gpus: IDs of GPUs to run inference on.--batch_size: Batch size of test loader.--path_to_images: Directory where images are stored.--test_list: Path to a file with image names and labels of the test data.--num_worker: Positive integer will turn on multi-process data loading with the specified number of loader worker processes (Check PyTorch docs).--checkpoint: Path to a checkpoint of the model (weights).--test_outdir: Directory where visualisation of the metrics in the pdf format will be saved.
The AUC score of the model on the test data is provided below.
The training algorithm is not deterministic, which means that the results may be slightly diffrenet even the model was trained on the same data and with the same hyperparameters.
If you use this code in research, please cite the following paper:
@misc{2003.08605,
Author = {Kudaibergen Urinbayev and Yerassyl Orazbek and Yernur Nurambek and Almas Mirzakhmetov and
Huseyin Atakan Varol},
Title = {End-to-End Deep Diagnosis of X-ray Images},
Year = {2020},
Eprint = {arXiv:2003.08605},
}