Image Annotation Incorporating Low-Rankness, Tag and Visual Correlation and Inhomogeneous Errors

TL;DR

This paper introduces a novel image annotation method that leverages low-rankness, tag-visual correlation, and inhomogeneous errors, using CNN features and semantic tag similarity to improve tag accuracy in image retrieval.

Contribution

The work proposes a new image annotation approach combining multiple priors and semantic tag similarity, solved efficiently with APG, outperforming existing methods.

Findings

Effective in handling incomplete and inaccurate tags

Demonstrates robustness across multiple datasets

Improves tag-visual correlation modeling

Abstract

Tag-based image retrieval (TBIR) has drawn much attention in recent years due to the explosive amount of digital images and crowdsourcing tags. However, TBIR is still suffering from the incomplete and inaccurate tags provided by users, posing a great challenge for tag-based image management applications. In this work, we proposed a novel method for image annotation, incorporating several priors: Low-Rankness, Tag and Visual Correlation and Inhomogeneous Errors. Highly representative CNN feature vectors are adopt to model the tag-visual correlation and narrow the semantic gap. And we extract word vectors for tags to measure similarity between tags in the semantic level, which is more accurate than traditional frequency-based or graph-based methods. We utilize the accelerated proximal gradient (APG) method to solve our model efficiently. Extensive experiments conducted on multiple benchmark datasets demonstrate the effectiveness and robustness of the proposed method.