Probabilistic Semi-Supervised Multi-Modal Hashing

Behnam Gholami and Abolfazl Hajisami

Abstract

Learning hash functions for high dimensional multi-modal data is of great interest for many real-world retrieval applications in which data comes from diverse heterogeneous sources. In this paper, we propose a novel probabilistic semi-supervised multi-modal retrieval model, by which we can learn both the binary codes and their dimension from the available training data. We also develop a new Variational Bayes (VB) algorithm for learning the parameters of the proposed model. The experiments on two real-world data sets show the superiority of the proposed method over other state-of-the-art algorithms for learning binary codes.

Session

Posters 1

Files

PDF iconExtended Abstract (PDF, 103K)
PDF iconPaper (PDF, 265K)

DOI

10.5244/C.30.28
https://dx.doi.org/10.5244/C.30.28

Citation

Behnam Gholami and Abolfazl Hajisami. Probabilistic Semi-Supervised Multi-Modal Hashing. In Richard C. Wilson, Edwin R. Hancock and William A. P. Smith, editors, Proceedings of the British Machine Vision Conference (BMVC), pages 28.1-28.12. BMVA Press, September 2016.

Bibtex

        @inproceedings{BMVC2016_28,
        	title={Probabilistic Semi-Supervised Multi-Modal Hashing},
        	author={Behnam Gholami and Abolfazl Hajisami},
        	year={2016},
        	month={September},
        	pages={28.1-28.12},
        	articleno={28},
        	numpages={12},
        	booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
        	publisher={BMVA Press},
        	editor={Richard C. Wilson, Edwin R. Hancock and William A. P. Smith},
        	doi={10.5244/C.30.28},
        	isbn={1-901725-59-6},
        	url={https://dx.doi.org/10.5244/C.30.28}
        }