Polyp Segmentation using Mutiple resolution strategy - HRSeg

In this project, we propose a novel method called HRSeg for polyp segmentation that uses multiple-resolution images to improve the accuracy of polyp segmentation. The proposed method outperforms the state-of-the-art methods for polyp segmentation, achieving higher about 3% accuracy.

Training strategy

Inference strategy

Pretrained model weight

Our HRSeg model weight can be download here GG drive

How to

Train

# Check arguments with
python train.py --help

Inference

python inference.py -n HRSeg -p path/to/model.pth

Analysis

python analysis.py -n1 HRSeg -n2 ssformer_S `
--print_table `
--show_scatter_dice_by_size `
--show_delta_dice `
--show_scatter_dice_by_range_size

Visualize

python visualize.py -n HRSeg -p path/to/model.pth