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

Thanks to the development of 2D keypoint detectors, monocular 3D human poseestimation (HPE) via 2D-to-3D uplifting approaches have achieved remarkableimprovements. Still, monocular 3D HPE is a challenging problem due to theinherent depth ambiguities and occlusions. To handle this problem, manyprevious works exploit temporal information to mitigate such difficulties.However, there are many real-world applications where frame sequences are notaccessible. This paper focuses on reconstructing a 3D pose from a single 2Dkeypoint detection. Rather than exploiting temporal information, we alleviatethe depth ambiguity by generating multiple 3D pose candidates which can bemapped to an identical 2D keypoint. We build a novel diffusion-based frameworkto effectively sample diverse 3D poses from an off-the-shelf 2D detector. Byconsidering the correlation between human joints by replacing the conventionaldenoising U-Net with graph convolutional network, our approach accomplishesfurther performance improvements. We evaluate our method on the widely adoptedHuman3.6M and HumanEva-I datasets. Comprehensive experiments are conducted toprove the efficacy of the proposed method, and they confirm that our modeloutperforms state-of-the-art multi-hypothesis 3D HPE methods.

DiffuPose: Monocular 3D Human Pose Estimation via Denoising Diffusion Probabilistic Model