This is the source code of paper "Contrastive Multi-bit Collaborative Learning for Deep Cross-modal Hashing".

• python 3.7.6
• pytorch 1.8.0
• torchvision 0.9.0
• numpy
• scipy
• tqdm
• pillow
• einops
• ftfy
• regex
• ...

Pretrained CLIP model could be found in the 30 lines of CLIP/clip/clip.py. This code is based on the "ViT-B/32". You should download "ViT-B/32" and put it in ./cache, or you can find it from the following link:

link: https://pan.baidu.com/s/1ZyDTR2IEHlY4xIdLgxtaVA password: kdq7

You should generate the following *.mat file for each dataset. The structure of directory ./dataset should be:

    dataset
    ├── coco
    │   ├── caption.mat 
    │   ├── index.mat
    │   └── label.mat 
    ├── flickr25k
    │   ├── caption.mat
    │   ├── index.mat
    │   └── label.mat
    └── nuswide
        ├── caption.mat
        ├── index.mat 
        └── label.mat

Please preprocessing dataset to appropriate input format.

More details about the generation, meaning, and format of each mat file could be found in ./dataset/README.md.

Additionally, cleaned datasets (MIRFLICKR25K & MSCOCO & NUSWIDE) used in our experiments are available at pan.baidu.com:

link: https://pan.baidu.com/s/1ZyDTR2IEHlY4xIdLgxtaVA password: kdq7

After preparing the python environment, pretrained CLIP model, and dataset, we can train the CMCL model.

See CMCL_$DATASET$.sh.

See CMCL_$DATASET$_TEST.sh.