| # ECG MCTnet — 5-class ECG Arrhythmia Classifier | |
| ## Model | |
| MCTnet: Dual-branch Conv + LTCAR Attention + VIB + Label Decoupling Head | |
| ## Classes | |
| - Normal | |
| - AFIB (Atrial Fibrillation) | |
| - AFL (Atrial Flutter) | |
| - LBBB (Left Bundle Branch Block) | |
| - RBBB (Right Bundle Branch Block) | |
| ## Performance | |
| - Val Macro-F1: 0.8198 | |
| - Dataset: TEKNOFEST ECG 2026 | |
| ## Input | |
| - Shape: (batch, 12, 5000) | |
| - 12-lead ECG @ 500Hz, 10 seconds | |
| ## Usage | |
| ```python | |
| import torch | |
| from huggingface_hub import hf_hub_download | |
| # Model indir | |
| weights = torch.load(hf_hub_download(repo_id="SarmAI/EKG-mctnet", filename="model.pt")) | |
| model = MCTnet(MCTnetConfig()) | |
| model.load_state_dict(weights) | |
| model.eval() | |
| ``` | |