Improved Search in Hamming Space using Deep Multi-Index Hashing

TL;DR

This paper introduces a deep multi-index hashing method that enhances search efficiency in Hamming space for large-scale image retrieval by learning balanced binary codes with improved speed and comparable accuracy.

Contribution

It proposes a novel deep-network-based multi-index hashing approach with balanced constraints to improve search efficiency in deep representation spaces.

Findings

Achieves significant speedups in image retrieval tasks.

Maintains comparable retrieval performance to existing methods.

Demonstrates effectiveness on multiple benchmark datasets.

Abstract

Similarity-preserving hashing is a widely-used method for nearest neighbour search in large-scale image retrieval tasks. There has been considerable research on generating efficient image representation via the deep-network-based hashing methods. However, the issue of efficient searching in the deep representation space remains largely unsolved. To this end, we propose a simple yet efficient deep-network-based multi-index hashing method for simultaneously learning the powerful image representation and the efficient searching. To achieve these two goals, we introduce the multi-index hashing (MIH) mechanism into the proposed deep architecture, which divides the binary codes into multiple substrings. Due to the non-uniformly distributed codes will result in inefficiency searching, we add the two balanced constraints at feature-level and instance-level, respectively. Extensive evaluations on several benchmark image retrieval datasets show that the learned balanced binary codes bring dramatic speedups and achieve comparable performance over the existing baselines.