English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Class activation maps are widely used for explaining deep neural networks.Due to its ability to highlight regions of interest, it has evolved in recentyears as a key step in weakly supervised learning. A major limitation to theperformance of the class activation maps is the small spatial resolution of thefeature maps in the last layer of the convolutional neural network. Therefore,we expect to generate high-resolution feature maps that result in high-qualitysemantic information. In this paper, we rethink the properties of semanticinformation in shallow feature maps. We find that the shallow feature mapsstill have fine-grained non-discriminative features while mixing considerablenon-target noise. Furthermore, we propose a simple gradient-based denoisingmethod to filter the noise by truncating the positive gradient. Our proposedscheme can be easily deployed in other CAM-related methods, facilitating thesemethods to obtain higher-quality class activation maps. We evaluate theproposed approach through a weakly-supervised semantic segmentation task, and alarge number of experiments demonstrate the effectiveness of our approach.

Rethinking Class Activation Maps for Segmentation: Revealing Semantic Information in Shallow Layers by Reducing Noise