Ontoloji-Based Generalized Zero-Shot Learning with Generative Networks

Abstract

Zero-Shot Learning (ZSL) aims to classify images of new categories in the testing phase without labeled images during training, using examples from categories with labeled images and some auxiliary information. The auxiliary information includes semantic attributes, textual descriptions, word embeddings, etc., for both labeled and unlabeled classes, utilizing Natural Language Processing (NLP) approaches. The word embeddings created are limited by the semantic attributes and textual descriptions where the semantics of categories are insufficient. In this paper, we introduce a study for Generalized Zero-Shot Learning (GZSL), a type of ZSL, by integrating the rich semantics offered by ontology. We support semantic representation using semantic attributes coupled with ontology. We employ Variational Autoencoder (VAE) and Generative Adversarial Network (GAN) architectures together to synthesize visual features. We evaluate our work on the AWA2 dataset and achieve improvements in GZSL performance.