Attend and Rectify: a Gated Attention Mechanism for Fine-Grained Recovery

TL;DR

This paper introduces a modular attention mechanism for CNNs that improves fine-grained recognition accuracy and robustness without needing part annotations, outperforming state-of-the-art methods across multiple datasets.

Contribution

A novel, architecture-independent attention mechanism that enhances CNNs for fine-grained recognition by learning to attend to lower-level features without part annotations.

Findings

Improves classification accuracy across multiple datasets.

Enhances robustness to clutter in images.

Surpasses state-of-the-art performance in several benchmarks.

Abstract

We propose a novel attention mechanism to enhance Convolutional Neural Networks for fine-grained recognition. It learns to attend to lower-level feature activations without requiring part annotations and uses these activations to update and rectify the output likelihood distribution. In contrast to other approaches, the proposed mechanism is modular, architecture-independent and efficient both in terms of parameters and computation required. Experiments show that networks augmented with our approach systematically improve their classification accuracy and become more robust to clutter. As a result, Wide Residual Networks augmented with our proposal surpasses the state of the art classification accuracies in CIFAR-10, the Adience gender recognition task, Stanford dogs, and UEC Food-100.