trends.barnas.net

task · first seen 7 Oct 2026

3D hand pose estimation

Also: hand pose estimation, 3D hand reconstruction, hand kinematics estimation, monocular hand mesh estimation

Estimating three-dimensional hand joint positions and motion from images or video.

3D hand pose estimation recovers the 3D positions of hand joints, and often a full hand mesh such as MANO, from images, video or other signals like EMG. It has been a standard computer vision task for about a decade. The sources show it being reworked for robot learning from human video. Papers from KAIST, Microsoft Research Asia and others build hand trackers on top of video foundation models such as V-JEPA, recover metric hand and camera trajectories from egocentric footage, and fuse vision with forearm sEMG. A fine-tuned SAM 3D Body hand branch (FAH) already has an Apple-silicon Metal port.

The task itself is long-established. The recent activity reflects its new role as a data source for dexterous robot policies, so better metric, physically consistent hand tracking from ordinary video could lower the cost of collecting manipulation demonstrations.

Analyst note (established), 18 Nov 2026.

Mentions per week

0219 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Rank 88 on 15 Nov 2026 · 8 mentions in 28 days · 8 independent groups · papers → code · first seen 6 weeks ago · acceleration +0.12

Earliest sources

Latest sources

Mentioned together with

surface electromyographyvideo diffusion modeltactile sensing estimationgenerative motion reconstructionjoint-embedding predictive architectureegocentric videogesture recognitionneural signal tokenizationhand-object interaction modelingMANO hand model