Towards Fast and Accurate Image-Text Retrieval with Self-Supervised Fine-Grained Alignment
This work addresses the problem of balancing speed and precision in cross-modal retrieval for applications like search engines, offering a solution that outperforms many interactive models with much lower computational cost.
The paper tackles the trade-off between accuracy and efficiency in image-text retrieval by introducing SelfAlign, a self-supervised fine-grained alignment module that improves retrieval accuracy by 9.1%, 4.2%, and 6.6% on benchmark datasets while maintaining fast retrieval times.
Image-text retrieval requires the system to bridge the heterogenous gap between vision and language for accurate retrieval while keeping the network lightweight-enough for efficient retrieval. Existing trade-off solutions mainly study from the view of incorporating cross-modal interactions with the independent-embedding framework or leveraging stronger pretrained encoders, which still demand time-consuming similarity measurement or heavyweight model structure in the retrieval stage. In this work, we propose an image-text alignment module SelfAlign on top of the independent-embedding framework, which improves the retrieval accuracy while maintains the retrieval efficiency without extra supervision. SelfAlign contains two collaborative sub-modules that force image-text alignment at both concept level and context level by self-supervised contrastive learning. It does not require cross-modal embedding interactions during training while maintaining independent image and text encoders during retrieval. With comparable time cost, SelfAlign consistently boosts the accuracy of state-of-the-art non-pretraining independent-embedding models respectively by 9.1%, 4.2% and 6.6% in terms of R@sum score on Flickr30K, MSCOCO 1K and MS-COCO 5K datasets. The retrieval accuracy also outperforms most existing interactive-embedding models with orders of magnitude decrease in retrieval time. The source code is available at: https://github.com/Zjamie813/SelfAlign.