TransHash: Transformer-based Hamming Hashing for Efficient Image Retrieval

Deep hamming hashing has gained growing popularity in approximate nearest neighbour search for large-scale image retrieval. Until now, the deep hashing for the image retrieval community has been dominated by convolutional neural network architectures, e.g. textttResnetcitehe2016deep. In this paper, inspired by the recent advancements of vision transformers, we present Transhash, a pure transformer-based framework for deep hashing learning. Concretely, our framework is composed of two major modules: (1) Based on textitVision Transformer (ViT), we design a siamese vision transformer backbone for image feature extraction. To learn fine-grained features, we innovate a dual-stream feature learning on top of the transformer to learn discriminative global and local features. (2) Besides, we adopt a Bayesian learning scheme with a dynamically constructed similarity matrix to learn compact binary hash codes. The entire framework is jointly trained in an end-to-end manner.~To the best of our knowledge, this is the first work to tackle deep hashing learning problems without convolutional neural networks (textitCNNs). We perform comprehensive experiments on three widely-studied datasets: CIFAR-10, NUSWIDE and IMAGENET. The experiments have evidenced our superiority against the existing state-of-the-art deep hashing methods. Specifically, we achieve 8.2%, 2.6%, 12.7% performance gains in terms of average textitmAP for different hash bit lengths on three public datasets, respectively.