Skip to content

Latest commit

 

History

History
68 lines (51 loc) · 1.67 KB

guidance.md

File metadata and controls

68 lines (51 loc) · 1.67 KB
Raw

Installation

a. Create a conda virtual environment and install required packages.

conda create -n mctrans pip python=3.7
conda activate mctrans
git clone https://github.com/JiYuanFeng/MCTrans.git
cd MCTrans
python setup.py develop
pip install -r requirements.txt

a. Complie other CUDA operators such as MultiScaleDeformableAttention.

cd mctrans/models/ops/
bash make.sh

c. Create data folder under the MCTrans and link the actual dataset path ($DATA_ROOT).

mkdir data
ln -s $DATA_ROOT data

Datasets Preparation

  • It is recommended to you to convert your dataset (espeacial the label) to standard format. For example, The binary segmengtaion label shoule only contain 0,1 or 0,255.

  • If your folder structure is different, you may need to change the corresponding paths in config files.

  • We have upload some preprocessed datasets at drive, you can download and unpack them under the data folder.

    MCTrans
├── mctrans
├── data
│   ├── pannuke
│   │   ├── split-images
│   │   ├── split-masks
│   │   ├── split-images-npy
│   │   ├── split-masks-npy
│   ├── cvc-clinic
│   │   ├── images
│   │   ├── masks
│   ├── cvc-colondb
│   │   ├── images
│   │   ├── masks
│   ├── kvasir
│   │   ├── images
│   │   ├── masks

Single GPU Training

bash tools/train.sh

Multi GPU Training

TO DO