From Captions to Visual Concepts and Back

Hao Fang, Saurabh Gupta, Forrest Iandola +9

This paper presents a novel approach for automatically generating image descriptions: visual detectors, language models, and multimodal similarity models learnt directly from a dat…