ouusan
commented
1 month ago 1.PAFs: Openpose https://arxiv.org/pdf/1812.08008 2.framework follow 2-45: Synthetic Training for Accurate 3D Huma Pose and Shape Estimation in the Wild https://github.com/akashsengupta1997/STRAPS-3DHumanShapePose. 3.Related works : Motion prior, Extrinsic camera calibration. different with 3-39 Character Controllers Using Motion VAEs https://arxiv.org/pdf/2103.14274
- RKTD inspired by KTD 4-32 Encoder-decoder with Multi-level Attention for 3D Human Shape and Pose Estimation related works: Foot-Ground Contact /3D Intermediate Feature Representation
1.Shape-aware Multi-Person Pose Estimation from Multi-View Images(2021) Part Affinity Fields (PAFs)，2d clusters candidate to 3d cluster candididate, confidence-aware votingbased algorithm. code: https://ait.ethz.ch/projects/2021/multi-human-pose/ 2.Probabilistic 3D Human Shape and Pose Estimation from Multiple Unconstrained Images in the Wild(2021) Shape distributions from each individual input are probabilistically combined to form a multi-input shape distribution. code:No 3.Dynamic Multi-Person Mesh Recovery From Uncalibrated Multi-View Cameras(2021) physics-geometry consistency(???), Latent Motion Prior.(VAE) code:No 4.Towards Stable Human Pose Estimation via Cross-View Fusion and Foot Stabilization(2023) a fused 3D intermediate representation by supervision over front, the side, and the top views, add foot-ground annotations to some public datasets, Reversible Kinematic Topology Decoder (RKTD) that adapt the reverse prediction order of the joints on the leg limb chain. code:No