Cross-Modal Implicit Relation Reasoning and Aligning for Text-to-Image Person Retrieval

CVPR 2023  ·  Ding Jiang, Mang Ye ·

Text-to-image person retrieval aims to identify the target person based on a given textual description query. The primary challenge is to learn the mapping of visual and textual modalities into a common latent space. Prior works have attempted to address this challenge by leveraging separately pre-trained unimodal models to extract visual and textual features. However, these approaches lack the necessary underlying alignment capabilities required to match multimodal data effectively. Besides, these works use prior information to explore explicit part alignments, which may lead to the distortion of intra-modality information. To alleviate these issues, we present IRRA: a cross-modal Implicit Relation Reasoning and Aligning framework that learns relations between local visual-textual tokens and enhances global image-text matching without requiring additional prior supervision. Specifically, we first design an Implicit Relation Reasoning module in a masked language modeling paradigm. This achieves cross-modal interaction by integrating the visual cues into the textual tokens with a cross-modal multimodal interaction encoder. Secondly, to globally align the visual and textual embeddings, Similarity Distribution Matching is proposed to minimize the KL divergence between image-text similarity distributions and the normalized label matching distributions. The proposed method achieves new state-of-the-art results on all three public datasets, with a notable margin of about 3%-9% for Rank-1 accuracy compared to prior methods.

PDF Abstract CVPR 2023 PDF CVPR 2023 Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Text-based Person Retrieval with Noisy Correspondence CUHK-PEDES IRRA Rank-1 69.74 # 3
Rank-5 87.09 # 2
Rank 10 92.20 # 2
mAP 62.28 # 3
mINP 45.84 # 3
Text based Person Retrieval CUHK-PEDES IRRA R@1 73.38 # 5
R@10 93.71 # 4
R@5 89.93 # 4
mAP 66.13 # 6
mINP 50.24 # 2
Text-based Person Retrieval with Noisy Correspondence ICFG-PEDES IRRA Rank 1 60.76 # 3
Rank-5 78.26 # 3
Rank-10 84.01 # 2
mAP 35.87 # 3
mINP 6.80 # 3
Text based Person Retrieval ICFG-PEDES IRRA mAP 38.06 # 6
R@1 63.46 # 6
R@5 80.25 # 5
R@10 85.82 # 3
mINP 7.93 # 1
Text-based Person Retrieval with Noisy Correspondence RSTPReid IRRA Rank 1 58.75 # 3
Rank 10 88.25 # 2
Rank 5 81.90 # 2
mAP 46.38 # 3
mINP 24.78 # 3
Text based Person Retrieval RSTPReid IRRA R@1 60.20 # 5
R@5 88.20 # 1
R@10 81.30 # 5

Methods