Structured Semantic Transfer for Multi-Label Recognition with Partial Labels

TL;DR

This paper introduces a structured semantic transfer framework that enables training multi-label image recognition models with only partial labels by leveraging intra-image and cross-image semantic correlations to generate pseudo labels.

Contribution

The proposed SST framework is the first to effectively utilize partial labels for multi-label recognition through dual semantic transfer modules, improving performance over existing methods.

Findings

Outperforms state-of-the-art algorithms on COCO, Visual Genome, and Pascal VOC datasets.

Effectively generates pseudo labels for unknown labels using semantic correlations.

Reduces annotation costs while maintaining high recognition accuracy.

Abstract

Multi-label image recognition is a fundamental yet practical task because real-world images inherently possess multiple semantic labels. However, it is difficult to collect large-scale multi-label annotations due to the complexity of both the input images and output label spaces. To reduce the annotation cost, we propose a structured semantic transfer (SST) framework that enables training multi-label recognition models with partial labels, i.e., merely some labels are known while other labels are missing (also called unknown labels) per image. The framework consists of two complementary transfer modules that explore within-image and cross-image semantic correlations to transfer knowledge of known labels to generate pseudo labels for unknown labels. Specifically, an intra-image semantic transfer module learns image-specific label co-occurrence matrix and maps the known labels to complement unknown labels based on this matrix. Meanwhile, a cross-image transfer module learns category-specific feature similarities and helps complement unknown labels with high similarities. Finally, both known and generated labels are used to train the multi-label recognition models. Extensive experiments on the Microsoft COCO, Visual Genome and Pascal VOC datasets show that the proposed SST framework obtains superior performance over current state-of-the-art algorithms. Codes are available at https://github.com/HCPLab-SYSU/HCP-MLR-PL.