Attributes2Classname: A discriminative model for attribute-based unsupervised zero-shot learning
This addresses the problem of reducing annotation labor and enabling text-only training for zero-shot learning, though it is incremental as it builds on existing attribute-based methods.
The paper tackles the difficulty of unsupervised zero-shot learning due to non-visual semantics in class name embeddings by discriminatively learning word representations that align class-attribute similarities with visual similarity, achieving state-of-the-art results on three benchmark datasets.
We propose a novel approach for unsupervised zero-shot learning (ZSL) of classes based on their names. Most existing unsupervised ZSL methods aim to learn a model for directly comparing image features and class names. However, this proves to be a difficult task due to dominance of non-visual semantics in underlying vector-space embeddings of class names. To address this issue, we discriminatively learn a word representation such that the similarities between class and combination of attribute names fall in line with the visual similarity. Contrary to the traditional zero-shot learning approaches that are built upon attribute presence, our approach bypasses the laborious attribute-class relation annotations for unseen classes. In addition, our proposed approach renders text-only training possible, hence, the training can be augmented without the need to collect additional image data. The experimental results show that our method yields state-of-the-art results for unsupervised ZSL in three benchmark datasets.