Semantic Diversity Learning for Zero-Shot Multi-label Classification

TL;DR

This paper introduces a novel zero-shot multi-label image classification model that leverages an embedding matrix with principal vectors and emphasizes semantic diversity during training, leading to state-of-the-art retrieval performance.

Contribution

It proposes an end-to-end training approach using an embedding matrix and a tailored loss function to better capture semantic diversity in multi-label zero-shot learning.

Findings

Achieves state-of-the-art results on NUS-Wide, COCO, and Open Images datasets.

Improves the ranking of relevant unseen labels in multi-label zero-shot classification.

Enhances image retrieval quality through semantic diversity modeling.

Abstract

Training a neural network model for recognizing multiple labels associated with an image, including identifying unseen labels, is challenging, especially for images that portray numerous semantically diverse labels. As challenging as this task is, it is an essential task to tackle since it represents many real-world cases, such as image retrieval of natural images. We argue that using a single embedding vector to represent an image, as commonly practiced, is not sufficient to rank both relevant seen and unseen labels accurately. This study introduces an end-to-end model training for multi-label zero-shot learning that supports semantic diversity of the images and labels. We propose to use an embedding matrix having principal embedding vectors trained using a tailored loss function. In addition, during training, we suggest up-weighting in the loss function image samples presenting higher semantic diversity to encourage the diversity of the embedding matrix. Extensive experiments show that our proposed method improves the zero-shot model's quality in tag-based image retrieval achieving SoTA results on several common datasets (NUS-Wide, COCO, Open Images).