Rank-Consistency Deep Hashing for Scalable Multi-Label Image Search

TL;DR

This paper introduces a novel deep hashing approach for scalable multi-label image retrieval that uses a rank-based supervision and a multi-label clustering loss to improve semantic consistency and discriminative power.

Contribution

The proposed method employs a rank-consistency objective and multi-label clustering loss, offering a new global supervision framework for multi-label hashing.

Findings

Achieves state-of-the-art results on MIRFLICKR-25K, IAPRTC12, and NUS-WIDE datasets.

Effectively aligns similarity orders between original and Hamming spaces.

Enhances discriminative power with a multi-label softmax cross-entropy loss.

Abstract

As hashing becomes an increasingly appealing technique for large-scale image retrieval, multi-label hashing is also attracting more attention for the ability to exploit multi-level semantic contents. In this paper, we propose a novel deep hashing method for scalable multi-label image search. Unlike existing approaches with conventional objectives such as contrast and triplet losses, we employ a rank list, rather than pairs or triplets, to provide sufficient global supervision information for all the samples. Specifically, a new rank-consistency objective is applied to align the similarity orders from two spaces, the original space and the hamming space. A powerful loss function is designed to penalize the samples whose semantic similarity and hamming distance are mismatched in two spaces. Besides, a multi-label softmax cross-entropy loss is presented to enhance the discriminative power with a concise formulation of the derivative function. In order to manipulate the neighborhood structure of the samples with different labels, we design a multi-label clustering loss to cluster the hashing vectors of the samples with the same labels by reducing the distances between the samples and their multiple corresponding class centers. The state-of-the-art experimental results achieved on three public multi-label datasets, MIRFLICKR-25K, IAPRTC12 and NUS-WIDE, demonstrate the effectiveness of the proposed method.