Hubness Reduction with Dual Bank Sinkhorn Normalization for Cross-Modal Retrieval (DBSN)
_{Official PyTorch implementation of the ACMMM 2025 paper}

Hubness Reduction with Dual Bank Sinkhorn Normalization for Cross-Modal Retrieval
Zhengxin Pan, Haishuai Wang, Fangyu Wu, Peng Zhang, Jiajun Bu
https://arxiv.org/abs/2508.02538

Requirements and Installation

We recommended the following dependencies.

• Python 3.7
• librosa
• PyTorch 1.11.0
• Transformers (4.18.0)

Results

Text-Image Retrieval

please refer to t2i.

Text-Video Retrieval

please refer to t2v.

Text-Audio Retrieval

please refer to t2a.

Image-Image Retrieval

please refer to i2i.

Text-Medical Retrieval

please refer to t2m.

Zero-shot Image Classification

please refer to ic.

Visualizations

please refer to vis.

Reference

If you found this code useful, please cite the following paper:

@inproceedings{pan2025dbsn,
    title={Hubness Reduction with Dual Bank Sinkhorn Normalization for Cross-Modal},
    author={Zhengxin Pan, Haishuai Wang, Fangyu Wu, Peng Zhang, Jiajun Bu},
    booktitle={Proceedings of the 33rd ACM International Conference on Multimedia (MM ’25), October 27–31, 2025, Dublin, Ireland},
    year={2025}
}