How rosedwayane/aieeg works
Project type and purpose
• This is a PyTorch-based machine-learning inference pipeline (no web frontend) for automatic EEG artifact removal.
• The README describes IC-U-Net and UNet++ variants, both U-Net-style denoising autoencoders for cleaning EEG signals stored as CSV.
File structure
• Top level:
• main.py – orchestration script for loading data, selecting model, running inference, and saving output.
• utils.py – data I/O, preprocessing/postprocessing helpers.
• requirements.txt – Python dependencies (numpy, scipy, torch, torchvision, torchaudio).
• sampledata/ – input CSV (30×timepoints) and placeholder output CSV.
• model/ – contains model definitions and checkpoints:
– ICUNet/ (architecture code + modelsave folder with .pth.tar checkpoints and train/val logs)
– UNetpp/ (similar structure)
– UNet_attention.py, UNet_family.py, cumbersome_model2.py – additional model variants.
• image/ – documentation asset (channel_location.JPG).
README for additional insights
• It confirms an end-to-end inference flow: set parameters in main.py (input path/name, sample rate, modelname, output path/name), then run python main.py.
• No explicit training script in repo (checkpoints are pre-trained and stored under model/*/modelsave).