Deep Convolutional Ranking for Multilabel Image Annotation

TL;DR

This paper introduces a deep convolutional ranking approach for multilabel image annotation, demonstrating that combining convolutional neural networks with approximate top-k ranking objectives significantly improves performance over traditional features.

Contribution

The work proposes a novel combination of convolutional architectures with approximate top-k ranking objectives for multilabel image annotation, achieving state-of-the-art results.

Findings

Outperforms conventional visual features by about 10% on NUS-WIDE dataset.

Combining CNN features with ranking objectives enhances multilabel annotation performance.

Achieves the best reported performance in the literature for the task.

Abstract

Multilabel image annotation is one of the most important challenges in computer vision with many real-world applications. While existing work usually use conventional visual features for multilabel annotation, features based on Deep Neural Networks have shown potential to significantly boost performance. In this work, we propose to leverage the advantage of such features and analyze key components that lead to better performances. Specifically, we show that a significant performance gain could be obtained by combining convolutional architectures with approximate top-$k$ ranking objectives, as thye naturally fit the multilabel tagging problem. Our experiments on the NUS-WIDE dataset outperforms the conventional visual features by about 10%, obtaining the best reported performance in the literature.