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

As natural images usually contain multiple objects, multi-label imageclassification is more applicable "in the wild" than single-labelclassification. However, exhaustively annotating images with every object ofinterest is costly and time-consuming. We aim to train multi-label classifiersfrom single-label annotations only. We show that adding a consistency loss,ensuring that the predictions of the network are consistent over consecutivetraining epochs, is a simple yet effective method to train multi-labelclassifiers in a weakly supervised setting. We further extend this approachspatially, by ensuring consistency of the spatial feature maps produced overconsecutive training epochs, maintaining per-class running-average heatmaps foreach training image. We show that this spatial consistency loss furtherimproves the multi-label mAP of the classifiers. In addition, we show that thismethod overcomes shortcomings of the "crop" data-augmentation by recoveringcorrect supervision signal even when most of the single ground truth object iscropped out of the input image by the data augmentation. We demonstrate gainsof the consistency and spatial consistency losses over the binary cross-entropybaseline, and over competing methods, on MS-COCO and Pascal VOC. We alsodemonstrate improved multi-label classification mAP on ImageNet-1K using theReaL multi-label validation set.

Spatial Consistency Loss for Training Multi-Label Classifiers from Single-Label Annotations