Multi-view Feature Augmentation with Adaptive Class Activation Mapping

TL;DR

This paper introduces an end-to-end trainable feature augmentation method for image classification that leverages multi-view local features and adaptive class activation mapping to improve model robustness and performance.

Contribution

It presents a novel multi-view feature augmentation module using AdaCAM for adaptive local feature sampling, enhancing classification accuracy over traditional global pooling methods.

Findings

Consistent performance improvements across experiments.

Effective sampling of class-discriminative local features.

Enhanced model robustness with multi-view augmentation.

Abstract

We propose an end-to-end-trainable feature augmentation module built for image classification that extracts and exploits multi-view local features to boost model performance. Different from using global average pooling (GAP) to extract vectorized features from only the global view, we propose to sample and ensemble diverse multi-view local features to improve model robustness. To sample class-representative local features, we incorporate a simple auxiliary classifier head (comprising only one 1$\times$1 convolutional layer) which efficiently and adaptively attends to class-discriminative local regions of feature maps via our proposed AdaCAM (Adaptive Class Activation Mapping). Extensive experiments demonstrate consistent and noticeable performance gains achieved by our multi-view feature augmentation module.