MSPNet Official Implementation of "MSPNet: Mutil-scale pooling learning for camouflaged instance segmentation"

Chen Li

[Paper]; [Project Page]

Contact: lichen_email@126.com

The code is tested on CUDA 11.1 and pytorch 1.9.0, change the versions below to your desired ones.

git clone https://github.com/another-u/MSPNet-main.git
cd MSPNet-main
conda create -n MSPNet python=3.8 -y
conda activate MSPNet
conda install pytorch==1.9.0 torchvision cudatoolkit=11.1 -c pytorch -c nvidia -y
python -m pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu111/torch1.9/index.html
python setup.py build develop

• Go to the link to download.

1. generate coco annotation files, you may refer to the tutorial of mmdetection for some help
2. change the path of the datasets as well as annotations in adet/data/datasets/cis.py, please refer to the docs of detectron2 for more help

# adet/data/datasets/cis.py
# change the paths 
DATASET_ROOT = 'Dataset_path'
ANN_ROOT = os.path.join(DATASET_ROOT, 'annotations')
TRAIN_PATH = os.path.join(DATASET_ROOT, 'Train/Image')
TEST_PATH = os.path.join(DATASET_ROOT, 'Test/Image')
TRAIN_JSON = os.path.join(ANN_ROOT, 'train_instance.json')
TEST_JSON = os.path.join(ANN_ROOT, 'test2026.json')
NC4K_ROOT = 'NC4K'
NC4K_PATH = os.path.join(NC4K_ROOT, 'Imgs')
NC4K_JSON = os.path.join(NC4K_ROOT, 'nc4k_test.json')

Model weights: P2T Weights.

The visual results are achieved by our MSPNet with P2T_tiny trained on the COD10K training set.

• Results on the COD10K test set: It will be updated afterwards.
• Results on the NC4K test set: It will be updated afterwards.

python tools/train_net.py --config-file configs/CIS_P2T.yaml --num-gpus 1 \
  OUTPUT_DIR {PATH_TO_OUTPUT_DIR}

Please replace {PATH_TO_OUTPUT_DIR} to your own output dir

python tools/train_net.py --config-file configs/CIS_P2T.yaml --eval-only \
  MODEL.WEIGHTS {PATH_TO_PRE_TRAINED_WEIGHTS}

Please replace {PATH_TO_PRE_TRAINED_WEIGHTS} to the pre-trained weights

python demo/demo.py --config-file configs/CIS_P2T.yaml \
  --input {PATH_TO_THE_IMG_DIR_OR_FIRE} \
  --output {PATH_TO_SAVE_DIR_OR_IMAGE_FILE} \
  --opts MODEL.WEIGHTS {PATH_TO_PRE_TRAINED_WEIGHTS}

• {PATH_TO_THE_IMG_DIR_OR_FIRE}: you can put image dir or image paths here
• {PATH_TO_SAVE_DIR_OR_IMAGE_FILE}: the place where the visualizations will be saved
• {PATH_TO_PRE_TRAINED_WEIGHTS}: please put the pre-trained weights here

If this helps you, please cite this work (MSPNet):

@article{li2024multi,
  title={Multi-scale pooling learning for camouflaged instance segmentation},
  author={Li, Chen and Jiao, Ge and Yue, Guowen and He, Rong and Huang, Jiayu},
  journal={Applied Intelligence},
  pages={1--15},
  year={2024},
  publisher={Springer}
}