Towards Better Guided Attention and Human Knowledge Insertion in Deep Convolutional Neural Networks

TL;DR

This paper introduces Multi-Scale Attention Branch Networks (MSABN) that improve attention resolution and performance in CNNs, and proposes a human knowledge insertion strategy using attention maps and bounding box annotations to enhance recognition tasks.

Contribution

The paper presents MSABN for better attention maps and a novel data augmentation method incorporating human annotations to boost CNN performance.

Findings

MSABN outperforms ABN and baseline models on benchmark datasets.

Attention map resolution is significantly improved with MSABN.

Human knowledge insertion via bounding boxes yields notable performance gains.

Abstract

Attention Branch Networks (ABNs) have been shown to simultaneously provide visual explanation and improve the performance of deep convolutional neural networks (CNNs). In this work, we introduce Multi-Scale Attention Branch Networks (MSABN), which enhance the resolution of the generated attention maps, and improve the performance. We evaluate MSABN on benchmark image recognition and fine-grained recognition datasets where we observe MSABN outperforms ABN and baseline models. We also introduce a new data augmentation strategy utilizing the attention maps to incorporate human knowledge in the form of bounding box annotations of the objects of interest. We show that even with a limited number of edited samples, a significant performance gain can be achieved with this strategy.