How hsd1503/resnet1d works
This repository provides PyTorch models for classifying one-dimensional signals, with ECG classification as a documented use case. In the PhysioNet workflow, preprocessing and dataset utilities prepare labeled, windowed signals; a selected backbone produces class scores, and the supplied checkpoint offers pretrained parameters. The README also documents synthetic-data exercise and optional Ray training/serving, but their implementations are not sampled, so those links are shown as unverified. The graph distinguishes dataset preparation from model architectures; no training orchestrator is inferred.
Main entry points: util.py Windowed datasets · resnet1d.py ResNet1D · net1d.py Net1D · acnn1d.py Attention CNN
Components 12
Signal data
- ECG preprocessing
util.py - Windowed datasets
util.py - Synthetic signals
util.py
Model zoo
- ResNet1D
resnet1d.py - Net1D
net1d.py - Attention CNN
acnn1d.py - CNN Transformer
cnn1d.py - CNN LSTM
crnn1d.py
Model assets
- Pretrained model
trained_model/model.pth
Serving
- Ray runtime
Other components
- PhysioNet ECG
- Model user
Connections 10
- PhysioNet ECG → ECG preprocessing: provides recordings
util.py - ECG preprocessing → Windowed datasets: prepares data
util.py - Windowed datasets → ResNet1D: model input
- Windowed datasets → Net1D: model input
- Windowed datasets → Attention CNN: model input
- Windowed datasets → CNN Transformer: model input
- Windowed datasets → CNN LSTM: model input
- Synthetic signals → Model user: generates examples
README.md - Model user → Pretrained model: loads parameters
README.md - Model user → Ray runtime: trains or serves
README.md