USER: Unified Semantic Enhancement With Momentum Contrast for Image-Text Retrieval

As a fundamental and challenging task in bridging language and vision domains, Image-Text Retrieval (ITR) aims at searching for the target instances that are semantically relevant to the given query from the other modality, and its key challenge is to measure the semantic similarity across different modalities. Although significant progress has been achieved, existing approaches typically suffer from two major limitations: (1) It hurts the accuracy of the representation by directly exploiting the bottom-up attention based region-level features where each region is equally treated. (2) It limits the scale of negative sample pairs by employing the mini-batch based end-to-end training mechanism. To address these limitations, we propose a Unified Semantic Enhancement Momentum Contrastive Learning (USER) method for ITR. Specifically, we delicately design two simple but effective Global representation based Semantic Enhancement (GSE) modules. One learns the global representation via the self-attention algorithm, noted as Self-Guided Enhancement (SGE) module. The other module benefits from the pre-trained CLIP module, which provides a novel scheme to exploit and transfer the knowledge from an off-the-shelf model, noted as CLIP-Guided Enhancement (CGE) module. Moreover, we incorporate the training mechanism of MoCo into ITR, in which two dynamic queues are employed to enrich and enlarge the scale of negative sample pairs. Meanwhile, a Unified Training Objective (UTO) is developed to learn from mini-batch based and dynamic queue based samples. Extensive experiments on the benchmark MSCOCO and Flickr30K datasets demonstrate the superiority of both retrieval accuracy and inference efficiency. For instance, compared with the existing best method NAAF, the metric R@1 of our USER on the MSCOCO 5K Testing set is improved by 5% and 2.4% on caption retrieval and image retrieval without any external knowledge or pre-trained model while enjoying over 60 times faster inference speed. Our source code will be released at https://github.com/zhangy0822/USER.

I. Introduction

Image-Text Retrieval (ITR) aims at searching for semantically relevant images from an image database for a given text query and vice versa. Nowadays, ITR has attracted extensive attention [1], [2], [3], [4], [5], [6], [7], [10] since there is an excellent benefit to extensive relevant applications, such as search engines and multimedia data management systems. The critical challenges of ITR lie in accurately obtaining visual and textual semantic representations, and learning a optimal joint embedding space from adequate sample pairs. Despite tremendous efforts have been dedicated to tackling the above challenges, such as the global-level matching methods [2], [6], [10], [11], attention-based local-level matching methods [1], [4], [12], [13], [14], external knowledge information based methods [3], [15], [16], [17], it remains challenging due to the inaccuracy of semantic representation and limited negative image-text pairs in a mini-batch.