Zero-Shot Learning Through Cross-Modal Transfer

TL;DR

This paper presents a zero-shot learning model that leverages unsupervised text data to recognize unseen objects in images, achieving state-of-the-art results without manual semantic features.

Contribution

It introduces a novel cross-modal transfer approach that uses language-based semantic information for image recognition, capable of handling both seen and unseen classes.

Findings

Achieves state-of-the-art performance on classes with many training images.

Performs reasonably well on unseen classes without training data.

Does not require manually defined semantic features.

Abstract

This work introduces a model that can recognize objects in images even if no training data is available for the objects. The only necessary knowledge about the unseen categories comes from unsupervised large text corpora. In our zero-shot framework distributional information in language can be seen as spanning a semantic basis for understanding what objects look like. Most previous zero-shot learning models can only differentiate between unseen classes. In contrast, our model can both obtain state of the art performance on classes that have thousands of training images and obtain reasonable performance on unseen classes. This is achieved by first using outlier detection in the semantic space and then two separate recognition models. Furthermore, our model does not require any manually defined semantic features for either words or images.