Weakly Supervised Localization Using Background Images

TL;DR

This paper introduces a novel end-to-end model that enlarges class activation maps for more precise object localization by leveraging background images as auxiliary supervision, achieving state-of-the-art results in weakly supervised localization.

Contribution

The work proposes a new method that uses background images to extract foreground regions, improving the localization accuracy of class activation maps in weakly supervised object localization.

Findings

Achieves state-of-the-art Top-1 and Top-5 localization error on CUB-200-2011.

Demonstrates competitive results on ILSVRC2016.

Enlarges CAMs for more accurate object localization.

Abstract

Weakly Supervised Object Localization (WSOL) methodsusually rely on fully convolutional networks in order to ob-tain class activation maps(CAMs) of targeted labels. How-ever, these networks always highlight the most discriminativeparts to perform the task, the located areas are much smallerthan entire targeted objects. In this work, we propose a novelend-to-end model to enlarge CAMs generated from classifi-cation models, which can localize targeted objects more pre-cisely. In detail, we add an additional module in traditionalclassification networks to extract foreground object propos-als from images without classifying them into specific cate-gories. Then we set these normalized regions as unrestrictedpixel-level mask supervision for the following classificationtask. We collect a set of images defined as Background ImageSet from the Internet. The number of them is much smallerthan the targeted dataset but surprisingly well supports themethod to extract foreground regions from different pictures.The region extracted is independent from classification task,where the extracted region in each image covers almost en-tire object rather than just a significant part. Therefore, theseregions can serve as masks to supervise the response mapgenerated from classification models to become larger andmore precise. The method achieves state-of-the-art results onCUB-200-2011 in terms of Top-1 and Top-5 localization er-ror while has a competitive result on ILSVRC2016 comparedwith other approaches.