Beyond Single Images: Retrieval Self-Augmented Unsupervised Camouflaged Object Detection
ICCV 2025
NKU & PolyU
- Create the Virtual Environment
conda create -n RISE python=3.8
conda activate RISE
pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 --extra-index-url https://download.pytorch.org/whl/cu113
git clone https://github.com/xiaohainku/RISE.git
cd RISE
pip install -r requirement.txt- Install FAISS package for efficient retrieval
Please refer to Installing FAISS for more details.
Install gpu version of faiss:
conda install -c pytorch -c nvidia faiss-gpu=1.7.3If you have difficulty in installing the gpu version of faiss, try to install the cpu verison:
pip install faiss-cpu==1.7.4If you have installed the cpu version, remember to change the "--faiss_device" argument in retrieval.py from "cuda" to "cpu".
Cluster maps, pseudo masks, and predicted maps are available at Google | Quark | Baidu.
RISE consists of two stages: generating pseudo masks and training SINet-V2 based on pseudo masks.
Step 1: Spectral clustering
python spectral-clustering.py \
--data_path /your/train/dataset/path \
--save_path /your/cluster/map/save/pathStep 2: Generate prototypes
python gen-proto.py \
--data_path /your/train/dataset/path \
--cluster_path /your/cluster/map/save/path \
--save_path /your/prototype/save/pathStep 3: Retrieval (generate pseudo mask)
python retrieval.py \
--data_path /your/train/dataset/path \
--prototype_path /your/prototype/save/path \
-save_path /your/pseudo/mask/save/pathStep 1: Training
cd SINet-V2
python MyTrain_Val.py \
--img_root /your/train/image/path/ \
--gt_root /your/pseudo/mask/path \
--val_root /your/validation/dataset/path/ \
--save_path /your/checkpoint/save/path/Step 2: Evaluation
python MyTesting.py \
--pth_path /your/checkpoint/save/path/ \
--data_path /your/test/dataset/path \
--save_dir /your/predicted/map/save/pathRISE builds upon self-supervised vision foundation model DINOv2. Thanks for their elegant work.
If you find our work interesting, please consider using the following BibTeX entry:
@InProceedings{RISE,
author = {Du, Ji and Wang, Xin and Hao, Fangwei and Yu, Mingyang and Chen, Chunyuan and Wu, Jiesheng and Wang, Bin and Xu, Jing and Li, Ping},
title = {Beyond Single Images: Retrieval Self-Augmented Unsupervised Camouflaged Object Detection},
booktitle = {IEEE International Conference on Computer Vision},
year = {2025},
pages = {22131-22142}
}