benchmarks

benchmarks of CinC and CPSC challenges, and using other databases

a large part are migrated from other DeepPSP repositories, some are implemented in the old fasion, being inconsistent with the new system architecture of torch_ecg, hence need updating and testing

Benchmark	Architecture	Source	Finished	Updated	Tested
CinC2020	CRNN	DeepPSP/cinc2020	✔️	✔️	✔️
CinC2021	CRNN	DeepPSP/cinc2021	✔️	✔️	✔️
CinC20221	Multi Task Learning (MTL)	DeepPSP/cinc2022	✔️	✔️	✔️
CPSC2019	SequenceTagging/U-Net	NA	✔️	✔️	✔️
CPSC2020	CRNN/SequenceTagging	DeepPSP/cpsc2020	✔️	❌	❌
CPSC2021	CRNN/SequenceTagging/LSTM	DeepPSP/cpsc2021	✔️	✔️	✔️
LUDB	U-Net	NA	✔️	✔️	✔️

Known Issues

1. Slicing data for CPSC2021 is too slow. An offline generated (sliced) dataset is hosted at Kaggle.
2. Dataset for LUDB is too slow
3. cli for training models are completely not tested

Footnotes

1. Although CinC2022 dealt with acoustic cardiac signals (phonocardiogram, PCG), the tasks and signals can be treated similarly. ↩