On Learning Where To Look

TL;DR

This paper introduces a learning-based visual recognition model that processes images through sequential glimpses, reducing computational costs and improving robustness to appearance variability, demonstrated on handwritten digit datasets.

Contribution

It presents a novel glimpse-based recognition model that scales with input complexity and learns parameters from data, addressing scalability and variability challenges.

Findings

Reduces computational complexity compared to full-image processing.

Demonstrates robustness to appearance changes.

Shows promising results on handwritten digit datasets.

Abstract

Current automatic vision systems face two major challenges: scalability and extreme variability of appearance. First, the computational time required to process an image typically scales linearly with the number of pixels in the image, therefore limiting the resolution of input images to thumbnail size. Second, variability in appearance and pose of the objects constitute a major hurdle for robust recognition and detection. In this work, we propose a model that makes baby steps towards addressing these challenges. We describe a learning based method that recognizes objects through a series of glimpses. This system performs an amount of computation that scales with the complexity of the input rather than its number of pixels. Moreover, the proposed method is potentially more robust to changes in appearance since its parameters are learned in a data driven manner. Preliminary experiments on a handwritten dataset of digits demonstrate the computational advantages of this approach.