SEINE is a framework for nuclei segmentation and reasoning in histopathological images. It supports training, evaluation, and interpretation of deep learning models for instance-level nuclei analysis.
Please refer to the detailed instructions in docs/data_prepare.md for dataset organization and formatting.
For preprocessing procedures such as patch extraction, data splitting, and structural encoding, please check docs/data.ipynb, which provides step-by-step processing code and explanations.
We recommend using a conda virtual environment.
# 1. Create environment
conda create -n seine python=3.7 -y
conda activate seine
# 2. Install PyTorch (ensure CUDA 11.1 support)
pip install torch==1.9.1+cu111 torchvision==0.10.1+cu111 torchaudio==0.9.1 \
-f https://download.pytorch.org/whl/torch_stable.html
# 3. Install MMCV-full (Linux recommended)
pip install mmcv-full==1.3.13
# 4. Install required packages
pip install -r requirements.txt
# 5. Clone and install this repo
git clone https://github.com/zhangye-zoe/SEINE.git
cd SEINE
pip install -e .✅ The code has been tested and runs successfully on both NVIDIA RTX 3090 and A100 GPUs.
To start training:
# Replace [config_path] with your actual config file
CUDA_VISIBLE_DEVICES=0 python tools/train.py [config_path]To evaluate a trained model:
# Replace [config_path] and [checkpoint] with your config file and checkpoint path
CUDA_VISIBLE_DEVICES=0 python tools/test.py [config_path] [checkpoint]SEINE/
├── configs/ # Configuration files
├── data/ # Dataset loading & interface
├── docs/ # Documentation and data preparation guides
├── tiseg/ # Core model architectures and components
├── tools/ # Training, testing, and utilities
├── requirements.txt # Python dependency list
└── README.md # Project overview
This project is inspired by the design pattern of
Tissue-Image-Segmentation. We thank their contribution to open-source medical image analysis.
For questions, please feel free to contact the author via GitHub Issues or email.