Vocabulary-informed Zero-shot and Open-set Learning

TL;DR

This paper introduces a vocabulary-informed learning framework that improves object recognition across supervised, zero-shot, and open-set scenarios by leveraging semantic distance constraints, enabling recognition with large vocabularies like ImageNet.

Contribution

It proposes a unified weighted maximum margin framework that incorporates vocabulary distance constraints for improved recognition in various settings, including large-scale open-set recognition.

Findings

Enhanced recognition accuracy in zero-shot and open-set tasks.

Effective handling of large vocabularies up to 310K classes.

Unified approach applicable across multiple recognition paradigms.

Abstract

Despite significant progress in object categorization, in recent years, a number of important challenges remain; mainly, the ability to learn from limited labeled data and to recognize object classes within large, potentially open, set of labels. Zero-shot learning is one way of addressing these challenges, but it has only been shown to work with limited sized class vocabularies and typically requires separation between supervised and unsupervised classes, allowing former to inform the latter but not vice versa. We propose the notion of vocabulary-informed learning to alleviate the above mentioned challenges and address problems of supervised, zero-shot, generalized zero-shot and open set recognition using a unified framework. Specifically, we propose a weighted maximum margin framework for semantic manifold-based recognition that incorporates distance constraints from (both supervised and unsupervised) vocabulary atoms. Distance constraints ensure that labeled samples are projected closer to their correct prototypes, in the embedding space, than to others. We illustrate that resulting model shows improvements in supervised, zero-shot, generalized zero-shot, and large open set recognition, with up to 310K class vocabulary on Animal with Attributes and ImageNet datasets.