Skip to content

Latest commit

 

History

History
15 lines (9 loc) · 997 Bytes

README.md

File metadata and controls

15 lines (9 loc) · 997 Bytes

this sub-folder is for the segmentation pipeline for our approach

This code requires Python 3.8.12 and PyTorch 1.8.2.

To use this code:

  • open a folder datasets in the root folder and keep Auto-retail-syndata-release inside it, see dataloader for details
  • run training/train_segmentation.py which trains U-Net
  • see test_segmentation.ipynb for making predictions and visualization on validation set using trained model
  • see test_segmentation.ipynb for making predictions and visualization on validation set using trained model and our post processing approach

Note on running test_segmentation.ipynb: either run your trained model or get pretrained weights from here. If you use the pretrained model, open a folder logs in root dir and keep it there.

The main output from this sub-folder is the *.pth file of the trained U-Net model which is to be used in the final testing stage.