Abstract

State-of-the-art methods for zero-shot visual recognition formulate learning as a joint embedding problem of images and side information. In these formulations the currentbest complement to visual features are attributes: manuallyencoded vectors describing shared characteristics among categories. Despite good performance, attributes have lim-itations: (1) finer-grained recognition requires commensu-rately more attributes, and (2) attributes do not provide a natural language interface. We propose to overcomethese limitations by training neural language models from scratch; i.e. without pre-training and only consuming words and characters. Our proposed models train end-to-end to align with the fine-grained and category-specific content ofimages. Natural language provides a flexible and compact way of encoding only the salient visual aspects for distinguishing categories. By training on raw text, our model can do inference on raw text as well, providing humans a familiar mode both for annotation and retrieval. Our model achieves strong performance on zero-shot text-based image retrieval and significantly outperforms the attribute-based state-of-the-art for zero-shot classification on the CaltechUCSD Birds 200-2011 dataset.