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

The backbone of traditional CNN classifier is generally considered as afeature extractor, followed by a linear layer which performs theclassification. We propose a novel loss function, termed as CAM-loss, toconstrain the embedded feature maps with the class activation maps (CAMs) whichindicate the spatially discriminative regions of an image for particularcategories. CAM-loss drives the backbone to express the features of targetcategory and suppress the features of non-target categories or background, soas to obtain more discriminative feature representations. It can be simplyapplied in any CNN architecture with neglectable additional parameters andcalculations. Experimental results show that CAM-loss is applicable to avariety of network structures and can be combined with mainstreamregularization methods to improve the performance of image classification. Thestrong generalization ability of CAM-loss is validated in the transfer learningand few shot learning tasks. Based on CAM-loss, we also propose a novelCAAM-CAM matching knowledge distillation method. This method directly uses theCAM generated by the teacher network to supervise the CAAM generated by thestudent network, which effectively improves the accuracy and convergence rateof the student network.

CAM-loss: Towards Learning Spatially Discriminative Feature Representations