Using Fictitious Class Representations to Boost Discriminative Zero-Shot Learners
This work addresses zero-shot learning for computer vision tasks, offering an incremental improvement to mitigate overfitting in seen classes.
The paper tackles the problem of discriminative zero-shot learning by introducing a mechanism that dynamically augments training with fictitious classes to reduce overfitting to attribute correlations in seen classes, resulting in improved state-of-the-art performance on the CUB dataset and comparable results on AWA2 and SUN datasets.
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.