Progressive Attention Networks for Visual Attribute Prediction

TL;DR

This paper introduces a progressive attention model for visual attribute prediction that iteratively refines focus on relevant image regions, improving accuracy over traditional methods.

Contribution

It presents a novel progressive attention mechanism that effectively attends to objects of various scales and shapes, incorporating local context for enhanced performance.

Findings

Outperforms traditional attention methods on synthetic and real datasets

Works well with hard attention mechanisms

Improves accuracy in visual attribute prediction tasks

Abstract

We propose a novel attention model that can accurately attends to target objects of various scales and shapes in images. The model is trained to gradually suppress irrelevant regions in an input image via a progressive attentive process over multiple layers of a convolutional neural network. The attentive process in each layer determines whether to pass or block features at certain spatial locations for use in the subsequent layers. The proposed progressive attention mechanism works well especially when combined with hard attention. We further employ local contexts to incorporate neighborhood features of each location and estimate a better attention probability map. The experiments on synthetic and real datasets show that the proposed attention networks outperform traditional attention methods in visual attribute prediction tasks.