Using Fictitious Class Representations to Boost Discriminative Zero-Shot Learners

TL;DR

This paper introduces a novel method for discriminative zero-shot learning that dynamically creates fictitious classes during training to improve generalization to unseen classes, achieving state-of-the-art results.

Contribution

The work proposes a dynamic fictitious class augmentation mechanism that reduces attribute correlation fixation in zero-shot learning models, enhancing their ability to generalize.

Findings

Improves state-of-the-art on CUB dataset

Achieves comparable results on AWA2 and SUN datasets

Analyzes effects of catastrophic forgetting in training

Abstract

Focusing on discriminative zero-shot learning, in this work we introduce a novel mechanism that dynamically augments during training the set of seen classes to produce additional fictitious classes. These fictitious classes diminish the model's tendency to fixate during training on attribute correlations that appear in the training set but will not appear in newly exposed classes. The proposed model is tested within the two formulations of the zero-shot learning framework; namely, generalized zero-shot learning (GZSL) and classical zero-shot learning (CZSL). Our model improves the state-of-the-art performance on the CUB dataset and reaches comparable results on the other common datasets, AWA2 and SUN. We investigate the strengths and weaknesses of our method, including the effects of catastrophic forgetting when training an end-to-end zero-shot model.