/AlignECG

Employing differentiable alignment to improve ECG classification.

Primary LanguageJupyter NotebookMIT LicenseMIT

AlignECG

Employing differentiable alignment to improve ECG classification.

example

Preparation

Place data file train.csv and test.csv under the ./data/ folder.

Install the corresponding dependencies:

pip install requirements.txt

Run the models

To run the model, using the

python main.py --model {model} \
        --log_path {log_path} \
        --ckpt_path {log_path} \
        --epoch {epoch_num}  \
        --lr {learning_rate} \
        --batch_size {batch_size} \
        --seed {random_seed} \
        --align {align_point_len} \
        --inference

  • model: The backbone model, select from [grufcn, trans, fcn, resnet, rescnn, gruatt, tcn]

  • align (opt.): The length of the aligned signal. If not set, no alignment is performed.

  • inference (opt.): Conduct inference over ./data/test.csv.

    For more details, please refer to run.sh and main.py.

Acknowledgements

Part of the codes is reused from tsai and DTAN.

License

Our code in this repository is licensed under the MIT license.