Semantic softmax loss for zero-shot learning

NEUROCOMPUTING

卷:316

Article

Ji, Zhong¹  Sun, Yuxin¹  Yu, Yunlong¹  Guo, Jichang¹  Pang, Yanwei¹ 

[1] Tianjin Univ, Sch Elect & Informat Engn, Tianjin, Peoples R China

关键词: Zero-shot learning;    Semantic embedding;    Multi-class classification;   

DOI  :  10.1016/j.neucom.2018.08.014

来源: Elsevier

【 摘 要 】

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 L-2 normalization constraint to the differences between the visual features and visual prototypes reconstructed with the semantic descriptors. The results on four benchmark datasets, i.e., AwA, CUB, SUN and ImageNet 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. (C) 2018 Elsevier B.V. All rights reserved.

【 授权许可】

Free