PCTrans: Position-Guided Transformer with Query Contrast for Biological Instance Segmentation (ICCVW 2023)

PCTrans is the first one-stage position-guided transformer with query contrast for biological instance segmentation.

Paper

This repository provides the official implementation of PCTrans.

PCTrans: Position-Guided Transformer with Query Contrast for Biological Instance Segmentation
Qi Chen¹, Wei Huang¹, Xiaoyu Liu¹, Jiacheng Li¹, Zhiwei Xiong^1
1 University of Science and Technology of China
paper | code | slides | poster

⏳ Dataset Link

• CVPPP
• BBBC

⏳ Command

cd connectomics/model/maskformer_block/pixel_decoder/ops
sh make.sh
cd ../../../../..

Training:

python -u -m torch.distributed.launch \
--nproc_per_node=4 --master_port=4321 scripts/main.py --distributed \
--config-base configs/CVPPP/CVPPP-PCTrans-Base.yaml \
--config-file configs/CVPPP/CVPPP-PCTrans.yaml

Inference:

python -u scripts/eval.py \
--config-base configs/CVPPP/CVPPP-PCTrans-Base.yaml \
--config-file configs/CVPPP/CVPPP-PCTrans.yaml --name cvppp

💡 Acknowledgement

This project is built upon previous projects. Especially, we'd like to thank the contributors of the following github repositories:

• Mask2Former
• pytorch_connectomics

🛡️ License

This project is under the CC-BY-NC 4.0 license. See LICENSE for details.

📦 Contact

[email protected]