Semantic Diversity Learning for Zero-Shot Multi-label Classification
This work addresses a real-world problem in image retrieval for natural images, but it is incremental as it builds on existing zero-shot learning approaches with specific improvements.
The paper tackles the challenge of zero-shot multi-label classification for images with numerous semantically diverse labels by proposing an end-to-end model that uses an embedding matrix and tailored loss functions to improve accuracy. The method achieves state-of-the-art results on datasets like NUS-Wide, COCO, and Open Images, enhancing tag-based image retrieval.
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).