# Semantic Softmax Loss for Zero-Shot Learning

## TL;DR

This paper introduces a nonlinear Semantic Softmax Loss for zero-shot learning that embeds class semantic descriptors into the classification layer, improving the alignment between visual features and semantic information.

## Contribution

It proposes a novel nonlinear approach with Semantic Softmax Loss and normalization constraints to enhance zero-shot learning performance.

## Key findings

- Achieves state-of-the-art results on AwA, CUB, and SUN datasets.
- Boosts zero-shot classification and retrieval performance.
- Demonstrates the effectiveness of nonlinear embedding in ZSL.

## Abstract

A typical pipeline for Zero-Shot Learning (ZSL) is to integrate the visual features and the class semantic descriptors into a multimodal framework with a linear or bilinear model. However, the visual features and the class semantic descriptors locate in different structural spaces, a linear or bilinear model can not capture the semantic interactions between different modalities well. In this letter, we propose a nonlinear approach to impose ZSL as a multi-class classification problem via a Semantic Softmax Loss by embedding the class semantic descriptors into the softmax layer of multi-class classification network. To narrow the structural differences between the visual features and semantic descriptors, we further use an L2 normalization constraint to the differences between the visual features and visual prototypes reconstructed with the semantic descriptors. The results on three benchmark datasets, i.e., AwA, CUB and SUN demonstrate the proposed approach can boost the performances steadily and achieve the state-of-the-art performance for both zero-shot classification and zero-shot retrieval.