🧊 NKU-Model-Compression

Official repository for model compression research projects from the Media Computing Lab, Nankai University.

This repository collects open-source projects on neural network compression, quantization, pruning, efficient inference, and deployment-oriented model optimization from MCG-NKU.

🌐 Explore the COD-TDQ Project Page →

A visual introduction to the W4A4 failure mechanism, DSTG and DCRP, quantitative and qualitative results, and measured deployment performance.

📦 Repository Notes

This is a code-first collection. Each project keeps its research implementation here, while dedicated project pages present the paper, results, datasets, and released checkpoints.

🚀 Projects

Project	Publication	Project Page	Venue
COD-TDQ code	When W4A4 Breaks Camouflaged Object Detection: Token-Group Dual-Constraint Activation Quantization	🌐 Open Website →	ECCV 2026

📚 Citations

If this repository is helpful, please consider giving it a star. If this project benefits your work, please consider citing the corresponding paper.

@inproceedings{li2026codtdq,
  title={When W4A4 Breaks Camouflaged Object Detection: Token-Group Dual-Constraint Activation Quantization},
  author={Li, Tianqi and Fang, Wenyu and He, Xin and Geng, Xue and Cheng, Xu and Liu, Yun},
  booktitle={Proceedings of the European Conference on Computer Vision},
  year={2026}
}

📬 Maintainers

This repository is maintained by the Media Computing Lab, Nankai University. New model compression projects will be added as independent subdirectories.

COD-TDQ contact: litianqi12 (3021210579@qq.com).