Research map / Video / Video generation
FunPhase: A Periodic Functional Autoencoder for Motion Generation via Phase Manifolds
Marco Pegoraro, Evan Atherton, Bruno Roy, Aliasghar Khani, Arianna Rampini
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| replace DeepPhase autoencoder with FunPhase functional periodic autoencoder | phase-manifold autoencoder · DOG reconstruction | position error (cm): 144 -> 61.4 | Improved |
| replace DeepPhase with FunPhase | phase-manifold autoencoder · DOG reconstruction | foot sliding: 1.76 -> 0.21 | Improved |
| add periodic phase decomposition to the functional autoencoder (ablation) | Function AE · 100STYLE reconstruction | foot sliding: 0.23 -> 0.16 (Function AE-256C -> FunPhase-32C) | Improved |
| add periodic phase decomposition to the functional autoencoder (ablation) | Function AE · 100STYLE reconstruction | NPSS: 0.59 (Function AE-256C) vs 0.75 (FunPhase-256C) | Mixed |
| increase latent channels from 32C to 256C (ablation) | FunPhase · 100STYLE reconstruction | position error (cm): 1.93 -> 0.36 | Improved |
| use ACMDM autoencoder instead of FunPhase | 100STYLE reconstruction · 100STYLE reconstruction | position error (cm): 0.32 (ACMDM-AE) vs 0.36 (FunPhase-256C); sliding 0.33 vs 0.16 | Mixed |
| add phase-based periodic latent component to functional diffusion (ablation) | Function Diffusion · 100STYLE latent diffusion | FID: 1.19 -> 0.51 | Improved |
| phase diffusion versus CAMDM on conditioning accuracy | FunPhase latent diffusion · 100STYLE latent diffusion | classification accuracy (%): 88.22 (CAMDM) vs 76.17 (FunPhase) | Hurt |
| use FunPhase latent diffusion instead of AnyTop | multi-skeleton generation · ZOO | coverage: 80 -> 96 | Improved |
Closest papers by meaning
- Neural Marionette: Unsupervised Learning of Motion Skeleton and Latent Dynamics from Volumetric Video (AAAI 2022 · similarity 0.87)
- Semantic-Aware Motion Encoding for Topology-Agnostic Character Animation (ICML 2026 · similarity 0.87)
- SAMoR: Motion Modelling for Articulated Objects of Any Skeleton and Topology (NeurIPS 2026 · similarity 0.86)
- LaDy: Lagrangian-Dynamic Informed Network for Skeleton-based Action Segmentation via Spatial-Temporal Modulation (CVPR 2026 · similarity 0.86)
- POMP: Physics-consistent Motion Generative Model through Phase Manifolds (CVPR 2025 · similarity 0.85)
- SurMo: Surface-based 4D Motion Modeling for Dynamic Human Rendering (CVPR 2024 · similarity 0.84)
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.
