# Learning Visually Consistent Label Embeddings for Zero-Shot Learning

## TL;DR

This paper introduces a zero-shot learning approach that aligns word vectors with visual features to improve recognition of unseen classes, demonstrating significant accuracy gains on benchmark datasets.

## Contribution

It presents a novel end-to-end method for jointly learning visually consistent word embeddings and label models for zero-shot learning.

## Key findings

- Significant improvement in recognition accuracy on benchmark datasets.
- Effective alignment of semantic and visual spaces for unseen classes.
- End-to-end training enhances zero-shot learning performance.

## Abstract

In this work, we propose a zero-shot learning method to effectively model knowledge transfer between classes via jointly learning visually consistent word vectors and label embedding model in an end-to-end manner. The main idea is to project the vector space word vectors of attributes and classes into the visual space such that word representations of semantically related classes become more closer, and use the projected vectors in the proposed embedding model to identify unseen classes. We evaluate the proposed approach on two benchmark datasets and the experimental results show that our method yields significant improvements in recognition accuracy.