Note
- This work has been accepted by ACMMM2025, and we are about to release all the code. 😊
- Details of the proposed Camoclass dataset can be found in the document for our dataset.
- Furthermore, we plan to provide more fine-grained textual annotations for the dataset. Regarding the training and testing environments, we aim to investigate the model’s capability to comprehend textual descriptions under complex environmental conditions.
- Additionally, we aspire that our precisely annotated categories will enable the research community to advance the exploration of camouflaged object semantic segmentation as a challenging yet promising task.
- We annotated the textual information for the key datasets DUTS, ECSSD, and PASCAL-S, DUTS.zip 提取码: ziat
To make access easier, we have provided several download links for the images, including Google Drive and a Data Repository. However, we do not hold the copyright to these images. It is your responsibility to review and adhere to the original licenses associated with the images before utilizing them.
The use of these images is entirely at your own risk and discretion. Additionally, please refer to the dataset description below for detailed instructions on how to set up Camoclass before executing our code.
- COD10K: https://github.com/GewelsJI/SINet-V2
- CAMO: https://sites.google.com/view/ltnghia/research/camo
- NC4K: https://github.com/xfflyer/Camouflaged-people-detection
- CHAMELEON: https://www.polsl.pl/rau6/chameleon-database-animal-camouflage-analysis
- Unzip all the files and set the following path information in a single yaml file
dataset.yaml:COD10K: Directory of COD10K, which containstrainandtestdirectories of COD10K.CAMO: Directory of CAMO, which containstrainandtestdirectories of CAMO.NC4K: Directory of NC4K, which containsimgandgtof NC4K.CHAMELEON: Directory of CHAMELEON, which containsimgandgtof CHAMELEON.
- Download for the information json for CamoClass dataset (
dataset_info.json) from - Specify the data roots for training and testing by dataset_info.json.
├── Camoclass
├── Camo
├── train # training set of camoclass
├── Imgs
├── Ant ...
└── test # tesing set of camoclass.
├── CAMO
├── Img ...
├── Ant ...
Referring Camouflaged Object Detection , TPAMI 2024.
Open-Vocabulary Camouflaged Object Segmentation , ECCV 2024.
Unlocking Attributes' Contribution to Successful Camouflage: A Combined Textual and Visual Analysis Strategy , ECCV 2024.
## 📎 Citation
If you find the code helpful in your research or work, please cite the following paper(s).
@article{zhang2024cgcod,
title={CGCOD: Class-Guided Camouflaged Object Detection},
author={Zhang, Chenxi and Zhang, Qing and Wu, Jiayun and Pang, Youwei},
journal={arXiv preprint arXiv:2412.18977},
year={2024}
}