Research map / 3D vision / Human pose estimation
Neural Kinematic Networks for Unsupervised Motion Retargetting
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add cycle consistency to the autoencoder objective (ablation) | Neural Kinematic Network · Mixamo online motion retargetting | MSE: 10.25 -> 8.51 | Helped |
| add adversarial training to cycle consistency (ablation) | Neural Kinematic Network · Mixamo online motion retargetting | MSE: 8.51 -> 7.10 | Helped |
| use the forward kinematics layer and recurrent architecture instead of a conditional RNN that outputs xyz coordinates | Neural Kinematic Network · Mixamo online motion retargetting | MSE: 13.65 -> 10.25 | Helped |
| add adversarial cycle consistency training to the conditional RNN baseline | Conditional RNN · Mixamo online motion retargetting | MSE: 13.65 -> 26.93 | Hurt |
| add adversarial cycle consistency training to the conditional MLP baseline | Conditional MLP · Mixamo online motion retargetting | MSE: 17.02 -> 16.96 | No clear effect |
| add adversarial training to cycle consistency (ablation) | Neural Kinematic Network · Known motion / known character | MSE: 5.68 -> 5.35 | Helped |
| add adversarial training to cycle consistency (ablation) | Neural Kinematic Network · Known motion / new character | MSE: 1.55 -> 1.35 | Helped |
| add adversarial training to cycle consistency (ablation) | Neural Kinematic Network · New motion / known character | MSE: 4.38 -> 4.39 | No clear effect |
| add adversarial training to cycle consistency (ablation) | Neural Kinematic Network · New motion / new character | MSE: 23.49 -> 18.02 | Helped |
| copy input joint rotations and global velocities instead of predicting them | Neural Kinematic Network · Mixamo online motion retargetting | MSE: 7.10 vs 9.00 | Mixed |
Closest papers by meaning
- Skinned Motion Retargeting With Residual Perception of Motion Semantics & Geometry (CVPR 2023 · similarity 0.88)
- MoCaNet: Motion Retargeting In-the-Wild via Canonicalization Networks (AAAI 2022 · similarity 0.88)
- Neural Marionette: Unsupervised Learning of Motion Skeleton and Latent Dynamics from Volumetric Video (AAAI 2022 · similarity 0.87)
- Flow Guided Transformable Bottleneck Networks for Motion Retargeting (CVPR 2021 · similarity 0.87)
- Semantics-aware Motion Retargeting with Vision-Language Models (CVPR 2024 · similarity 0.85)
- REACTO: Reconstructing Articulated Objects from a Single Video (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.