🦎 COD Dataset Bundle For COD-TDQ

This repository hosts the camouflaged object detection dataset bundle used by:

When W4A4 Breaks Camouflaged Object Detection: Token-Group Dual-Constraint Activation Quantization

• Paper: https://arxiv.org/abs/2604.16855
• Code: https://github.com/MCG-NKU/nku-model-compre/tree/main/cod-tdq

This dataset repository contains the COD resource bundle used by the release: TrainValDataset plus CHAMELEON, CAMO, COD10K and NC4K test sets.

Layout

Dataset/
├── TrainValDataset/{Imgs,GT,Edge}
└── TestDataset/
    ├── CHAMELEON/{Imgs,GT,Edge}
    ├── CAMO/{Imgs,GT,Edge}
    ├── COD10K/{Imgs,GT,Edge}
    └── NC4K/{Imgs,GT,Instance}

Counts

Split / Dataset	Images
TrainValDataset	4040
CHAMELEON	76
CAMO	250
COD10K	2026
NC4K	4121

Usage

Use this dataset with the official code:

git clone https://github.com/MCG-NKU/nku-model-compre.git
cd nku-model-compre/cod-tdq

Expected local layout:

assets/datasets/cod/
├── TrainValDataset/{Imgs,GT,Edge}
└── TestDataset/{CHAMELEON,CAMO,COD10K,NC4K}

If you download this repository directly, move or symlink Dataset/TrainValDataset and Dataset/TestDataset into assets/datasets/cod/.

Citation

@inproceedings{codtdq2026,
  title={When W4A4 Breaks Camouflaged Object Detection: Token-Group Dual-Constraint Activation Quantization},
  author={Coming Soon},
  booktitle={ECCV},
  year={2026}
}

Please also cite the original datasets according to their respective licenses.

Integrity

Checksums are provided in SHA256SUMS.