SCH

Source code for TPAMI'24 paper "Cross-Modal Hashing Method with Properties of Hamming Space: A New Perspective"

You may find our other works on CMH (Paper & Code):

• [TCSVT'23] Joint Semantic Preserving Sparse Hashing for Cross-Modal Retrieval
• [ICASSP'24] Key Points Centered Sparse Hashing for Cross-Modal Retrieval
• [ICMR'19] Triplet Fusion Network Hashing for Unpaired Cross-Modal Retrieval
• [TPAMI'19] MTFH: A Matrix Tri-Factorization Hashing Framework for Efficient Cross-Modal Retrieval

Datasets

Please refer to the provided link to download the dataset, create a data folder and update data path in settings.py.

Train model

You can directly run the file

python train.py --Bit 16 --GID 0 --DS 0

to get the results.

Evaluate the model

Modify the settings.py line 7

EVAL = True

You can downlod the trained models via following links (have been updated):

Dataset	Hash Bit	Downlod
MIR	16	link
MIR	32	link
MIR	64	link
MIR	128	link
NUS	16	link
NUS	32	link
NUS	64	link
NUS	128	link

Citation

If you find SCH useful in your research, please consider citing:

@article{hu2024cross,
  title={Cross-Modal Hashing Method with Properties of Hamming Space: A New Perspective},
  author={Hu, Zhikai and Cheung, Yiu-ming and Li, Mengke and Lan, Weichao},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2024},
  volume={46},
  number={12},
  pages={7636-7650},
  doi={10.1109/TPAMI.2024.3392763}}
}