Recurrent Models of Visual Attention

TL;DR

This paper introduces a recurrent neural network model that adaptively focuses on specific image regions, reducing computation for large images and improving performance on cluttered image classification and dynamic visual tasks.

Contribution

The paper presents a novel recurrent attention model that selectively processes image regions, enabling efficient computation and learning task-specific focus through reinforcement learning.

Findings

Outperforms CNN baseline on cluttered image classification

Successfully tracks objects without explicit training signals

Reduces computation independently of image size

Abstract

Applying convolutional neural networks to large images is computationally expensive because the amount of computation scales linearly with the number of image pixels. We present a novel recurrent neural network model that is capable of extracting information from an image or video by adaptively selecting a sequence of regions or locations and only processing the selected regions at high resolution. Like convolutional neural networks, the proposed model has a degree of translation invariance built-in, but the amount of computation it performs can be controlled independently of the input image size. While the model is non-differentiable, it can be trained using reinforcement learning methods to learn task-specific policies. We evaluate our model on several image classification tasks, where it significantly outperforms a convolutional neural network baseline on cluttered images, and on a dynamic visual control problem, where it learns to track a simple object without an explicit training signal for doing so.