Fine-grained Visual Textual Alignment for Cross-Modal Retrieval using Transformer Encoders

TL;DR

This paper introduces TERAN, a transformer-based model that achieves state-of-the-art fine-grained image-sentence alignment for cross-modal retrieval, emphasizing scalable and efficient retrieval pipelines.

Contribution

The paper proposes a novel transformer encoder approach that enforces fine-grained image-text alignment while maintaining separate data pipelines for scalable retrieval.

Findings

State-of-the-art results on MS-COCO and Flickr30k datasets.

Improved Recall@1 by 5.7% for image retrieval on MS-COCO.

Enhanced sentence retrieval performance with a 3.5% increase in Recall@1.

Abstract

Despite the evolution of deep-learning-based visual-textual processing systems, precise multi-modal matching remains a challenging task. In this work, we tackle the task of cross-modal retrieval through image-sentence matching based on word-region alignments, using supervision only at the global image-sentence level. Specifically, we present a novel approach called Transformer Encoder Reasoning and Alignment Network (TERAN). TERAN enforces a fine-grained match between the underlying components of images and sentences, i.e., image regions and words, respectively, in order to preserve the informative richness of both modalities. TERAN obtains state-of-the-art results on the image retrieval task on both MS-COCO and Flickr30k datasets. Moreover, on MS-COCO, it also outperforms current approaches on the sentence retrieval task. Focusing on scalable cross-modal information retrieval, TERAN is designed to keep the visual and textual data pipelines well separated. Cross-attention links invalidate any chance to separately extract visual and textual features needed for the online search and the offline indexing steps in large-scale retrieval systems. In this respect, TERAN merges the information from the two domains only during the final alignment phase, immediately before the loss computation. We argue that the fine-grained alignments produced by TERAN pave the way towards the research for effective and efficient methods for large-scale cross-modal information retrieval. We compare the effectiveness of our approach against relevant state-of-the-art methods. On the MS-COCO 1K test set, we obtain an improvement of 5.7% and 3.5% respectively on the image and the sentence retrieval tasks on the Recall@1 metric. The code used for the experiments is publicly available on GitHub at https://github.com/mesnico/TERAN.