Areas of Attention for Image Captioning

TL;DR

This paper introduces a novel attention-based model for image captioning that models dependencies between image regions, caption words, and RNN states, leading to improved localization and state-of-the-art results.

Contribution

It proposes a new attention mechanism that directly associates caption words with image regions and compares different methods for generating attention areas, including spatial transformers.

Findings

Spatial transformers outperform other methods for attention areas.

The model achieves state-of-the-art results on MSCOCO.

Associating caption words with image regions improves caption accuracy.

Abstract

We propose "Areas of Attention", a novel attention-based model for automatic image captioning. Our approach models the dependencies between image regions, caption words, and the state of an RNN language model, using three pairwise interactions. In contrast to previous attention-based approaches that associate image regions only to the RNN state, our method allows a direct association between caption words and image regions. During training these associations are inferred from image-level captions, akin to weakly-supervised object detector training. These associations help to improve captioning by localizing the corresponding regions during testing. We also propose and compare different ways of generating attention areas: CNN activation grids, object proposals, and spatial transformers nets applied in a convolutional fashion. Spatial transformers give the best results. They allow for image specific attention areas, and can be trained jointly with the rest of the network. Our attention mechanism and spatial transformer attention areas together yield state-of-the-art results on the MSCOCO dataset.o meaningful latent semantic structure in the generated captions.