Compositional Correlation Quantization for Large-Scale Multimodal Search

Explore this paper's citation graph

Summary

This paper establishes seamless multimodal hashing by proposing a novel Compositional Correlation Quantization (CCQ) model, which jointly finds correlation-maximal mappings that transform different modalities into an isomorphic latent space, and learns compositional quantizers that quantize the isomorph latent features into compact binary codes.

Type
preprint
Published
2015-04-19
Cited by
5
References
38
Access
Open access

Keywords

Computer science, Nearest neighbor search, Hash function, Quantization (signal processing), Binary code

References

Cited by

Related papers