Deep Reinforcement Learning with Label Embedding Reward for Supervised Image Hashing

TL;DR

This paper introduces a novel deep reinforcement learning approach for supervised image hashing, using label embedding rewards to improve binary code generation, leading to better retrieval performance.

Contribution

It proposes a decision-making framework with a deep Q-network and BCH code-based rewards, which is a new approach in supervised image hashing.

Findings

Outperforms state-of-the-art methods on CIFAR-10 and NUS-WIDE datasets.

Effective exploration of binary code space via reinforcement learning.

Improved retrieval accuracy with various code lengths.

Abstract

Deep hashing has shown promising results in image retrieval and recognition. Despite its success, most existing deep hashing approaches are rather similar: either multi-layer perceptron or CNN is applied to extract image feature, followed by different binarization activation functions such as sigmoid, tanh or autoencoder to generate binary code. In this work, we introduce a novel decision-making approach for deep supervised hashing. We formulate the hashing problem as travelling across the vertices in the binary code space, and learn a deep Q-network with a novel label embedding reward defined by Bose-Chaudhuri-Hocquenghem (BCH) codes to explore the best path. Extensive experiments and analysis on the CIFAR-10 and NUS-WIDE dataset show that our approach outperforms state-of-the-art supervised hashing methods under various code lengths.