Polyp Semi-supervised segmentation with Online Pseudo labeling and Momentum network

Original code for paper "Online pseudo labeling for polyp segmentation with momentum networks"

How to use this source code

Requirement

• Python 3.7
• CUDA 11.1
• Wandb account for visualization. Register at https://wandb.ai/

Installation

pip install -r requirements.txt 

Data Prepare

Dataset training including Kvasir-SEG, CVC-ColonDB, EndoScene, ETIS-Larib Polyp DB and CVC-Clinic DB from PraNet: Parallel Reverse Attention Network for Polyp Segmentation

• Download training dataset and move it into your ./data folder which can be found in this Google Drive
• Download testing dataset and move it into your ./data folder which can be found in this Google Drive

Training

After download all data and setting hyper params. We run this command to train models

python train.py

Testing

• Download pretrained model from this Google Drive and put all the models downloaded to the /pretrained folder.
• Change the relative path of model in test.py file.
• Run test script

python test.py

Visualization

for test samples visualization, we use wandb website. You can use online service at https://wandb.ai/