LACFormer: Toward accurate and efficient polyp segmentation

This repository contains the official Pytorch implementation of training & evaluation code for LACFormer.

Environment

• Creating a virtual environment in terminal: conda create -n LACFormer python=3.8.16
• Install CUDA 11.3 and pytorch 1.8.1: conda install pytorch==1.8.1 torchvision==0.9.1 torchaudio==0.8.1 cudatoolkit=11.3 -c pytorch -c conda-forge
• Install other requirements: pip install -r requirements.txt

Dataset

Downloading necessary data: For Experiment in our paper:

• Download testing dataset, move it into Dataset/ and extract the file zip, which can be found in this download link (Google Drive).
• Download training dataset, move it into Dataset/ and extract the file zip, which can be found in this download link (Google Drive).

Training

Download MiT's pretrained weights on ImageNet-1K, and put them in a folder pretrained/. Config hyper-parameters in mcode/config.py and run train.py for training:

python train.py

Here is an example in Google Colab

Evaluation

After training, evaluation will be done automatically

Checkpoint

The checkpoint for LACFormer-L can be downloaded from here