Monocular video method combining SAM 3D Body with biomechanical inverse kinematics achieves approximately 10 degree finger joint angle error and 6 mm hand position error versus multiview reconstruction on 4590 frames.
Equinox: neural networks in JAX via callable PyTrees and filtered transformations
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Monocular Biomechanical Tracking of Fingers with Inverse Kinematics to Foundation Models
Monocular video method combining SAM 3D Body with biomechanical inverse kinematics achieves approximately 10 degree finger joint angle error and 6 mm hand position error versus multiview reconstruction on 4590 frames.