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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

DensePose supersedes traditional landmark detectors by densely mapping imagepixels to body surface coordinates. This power, however, comes at a greatlyincreased annotation time, as supervising the model requires to manually labelhundreds of points per pose instance. In this work, we thus seek methods tosignificantly slim down the DensePose annotations, proposing more efficientdata collection strategies. In particular, we demonstrate that if annotationsare collected in video frames, their efficacy can be multiplied for free byusing motion cues. To explore this idea, we introduce DensePose-Track, adataset of videos where selected frames are annotated in the traditionalDensePose manner. Then, building on geometric properties of the DensePosemapping, we use the video dynamic to propagate ground-truth annotations in timeas well as to learn from Siamese equivariance constraints. Having performedexhaustive empirical evaluation of various data annotation and learningstrategies, we demonstrate that doing so can deliver significantly improvedpose estimation results over strong baselines. However, despite what issuggested by some recent works, we show that merely synthesizing motionpatterns by applying geometric transformations to isolated frames issignificantly less effective, and that motion cues help much more when they areextracted from videos.

Slim DensePose: Thrifty Learning from Sparse Annotations and Motion Cues