An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval

13 Jun 2024  ·  Jaeseok Byun, Seokhyeon Jeong, Wonjae Kim, Sanghyuk Chun, Taesup Moon ·

Composed Image Retrieval (CIR) aims to retrieve a target image based on a reference image and conditioning text, enabling controllable image searches. The mainstream Zero-Shot (ZS) CIR methods bypass the need for expensive training CIR triplets by projecting image embeddings into the text token embedding space, forming a composed query for retrieval. However, we highlight an inherent limitation in these projection-based CIR: a task discrepancy of text encoders between the original pre-training task of the encoders (text $\leftrightarrow$ image) and the target CIR task (image + text $\leftrightarrow$ image), which potentially negatively impacts CIR performance. To reduce such a discrepancy, a naive solution would be to train both image and text encoders with CIR triplets in a supervised manner. Instead, we introduce Reducing Task Discrepancy of Text Encoders (RTD), an efficient text-only post-hoc framework that complements projection-based CIR methods. We devise a novel target-anchored text contrastive learning designed to enhance the capability of the text encoder for CIR. We also propose two key enhancements: (1) a hard negative-based refined batch sampling strategy and (2) a refined concatenation scheme to further mitigate training-inference discrepancy. Integrating RTD into state-of-the-art projection-based methods achieves performance comparable to, or even surpassing, resource-intensive state-of-the-art synthetic CIR triplet-based approaches only with 23 minutes of additional training on 4 A100 GPUs (up to $100\times$ faster in training). Our code will be available upon acceptance.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Zero-Shot Composed Image Retrieval (ZS-CIR) CIRCO RTD + LinCIR (CLIP L/14) mAP@10 18.11 # 27
Zero-Shot Composed Image Retrieval (ZS-CIR) CIRCO RTD + LinCIR (CLIP G/14) mAP@10 22.29 # 20
Zero-Shot Composed Image Retrieval (ZS-CIR) CIRR RTD + LinCIR (CLIP G/14) R@5 67.47 # 6
Zero-Shot Composed Image Retrieval (ZS-CIR) CIRR RTD + LinCIR (CLIP L/14) R@5 56.17 # 26
Zero-Shot Composed Image Retrieval (ZS-CIR) Fashion IQ RTD + LinCIR (CLIP G/14) (Recall@10+Recall@50)/2 56.74 # 1
Zero-Shot Composed Image Retrieval (ZS-CIR) Fashion IQ RTD + LinCIR (CLIP L/14) (Recall@10+Recall@50)/2 40.66 # 22

Methods