From captions to visual concepts and back

TLDR

The paper proposes a novel method that learns visual detectors, a maximum‑entropy language model, and a multimodal similarity model directly from image captions to automatically generate image descriptions. It trains word‑level visual detectors with multiple instance learning, feeds their outputs into a language model trained on 400,000 captions, and re‑ranks generated captions using sentence‑level features and a deep multimodal similarity model to capture global semantics. The system attains state‑of‑the‑art results on the Microsoft COCO benchmark with a BLEU‑4 score of 29.1%, and human judges find its captions equal or better than human‑written ones 34% of the time.

Abstract

This paper presents a novel approach for automatically generating image descriptions: visual detectors, language models, and multimodal similarity models learnt directly from a dataset of image captions. We use multiple instance learning to train visual detectors for words that commonly occur in captions, including many different parts of speech such as nouns, verbs, and adjectives. The word detector outputs serve as conditional inputs to a maximum-entropy language model. The language model learns from a set of over 400,000 image descriptions to capture the statistics of word usage. We capture global semantics by re-ranking caption candidates using sentence-level features and a deep multimodal similarity model. Our system is state-of-the-art on the official Microsoft COCO benchmark, producing a BLEU-4 score of 29.1%. When human judges compare the system captions to ones written by other people on our held-out test set, the system captions have equal or better quality 34% of the time.