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

Recovering 3D human mesh from monocular images is a popular topic in computervision and has a wide range of applications. This paper aims to estimate 3Dmesh of multiple body parts (e.g., body, hands) with large-scale differencesfrom a single RGB image. Existing methods are mostly based on iterativeoptimization, which is very time-consuming. We propose to train a single-shotmodel to achieve this goal. The main challenge is lacking training data thathave complete 3D annotations of all body parts in 2D images. To solve thisproblem, we design a multi-branch framework to disentangle the regression ofdifferent body properties, enabling us to separate each component's training ina synthetic training manner using unpaired data available. Besides, tostrengthen the generalization ability, most existing methods have usedin-the-wild 2D pose datasets to supervise the estimated 3D pose via 3D-to-2Dprojection. However, we observe that the commonly used weak-perspective modelperforms poorly in dealing with the external foreshortening effect of cameraprojection. Therefore, we propose a depth-to-scale (D2S) projection toincorporate the depth difference into the projection function to deriveper-joint scale variants for more proper supervision. The proposed methodoutperforms previous methods on the CMU Panoptic Studio dataset according tothe evaluation results and achieves comparable results on the Human3.6M bodyand STB hand benchmarks. More impressively, the performance in close shotimages gets significantly improved using the proposed D2S projection for weaksupervision, while maintains obvious superiority in computational efficiency.

Synthetic Training for Monocular Human Mesh Recovery