Research map / Video / Video generation
Tracking-Guided 4D Generation: Foundation-Tracker Motion Priors for 3D Model Animation
Su Sun, Cheng Zhao, Himangi Mittal, Gaurav Mittal, Rohith Kukkala, Yingjie Victor Chen, Mei Chen
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 |
|---|---|---|---|
| add correspondence tracking loss from CoTracker3 to diffusion features (ablation) | Track4DGen stage one · Diffusion4D filtered, multi-view video generation | I2V: 0.844 -> 0.933 | Improved |
| add correspondence tracking loss (ablation) | Track4DGen stage one · Diffusion4D filtered, multi-view video generation | Aesthetic Quality: 0.347 -> 0.470 | Improved |
| add correspondence tracking loss (ablation) | Track4DGen stage one · Diffusion4D filtered, multi-view video generation | Dynamic Score: 1.505 -> 1.356 | Hurt |
| add position loss (ablation) | Track4DGen stage one · Diffusion4D filtered, multi-view video generation | I2V: 0.921 -> 0.933 | Improved |
| add position loss (ablation) | Track4DGen stage one · Diffusion4D filtered, multi-view video generation | Dynamic Score: 1.435 -> 1.356 | Hurt |
| add Stage One diffusion features to Hex-plane features in 4D-GS (ablation) | Track4DGen stage two · Sketchfab28 4D generation ablation | Aesthetic Quality: 0.532 -> 0.538 | Improved |
| add 4D spherical harmonics color model (ablation) | Track4DGen stage two · Sketchfab28 4D generation ablation | Aesthetic Quality: 0.536 -> 0.538 | Improved |
| add diffusion features to hybrid motion representation (ablation) | Track4DGen stage two · Sketchfab28 4D generation ablation | Dynamic Score: 1.292 -> 1.286 | No clear effect |
| use Track4DGen instead of Animate3D | multi-view video generation · Diffusion4D filtered, multi-view video generation | I2V: 0.919 -> 0.933 | Improved |
| use Track4DGen instead of Animate3D | multi-view video generation · Animate3D dataset | Dynamic Score: 0.787 -> 0.778 | Hurt |
| use Track4DGen instead of Animate3D | 4D generation · Sketchfab28 | CLIP-O(img): 0.8812 -> 0.8884 | Improved |
| use Track4DGen instead of Animate3D | 4D generation · Animate3D dataset | CLIP-F(img): 0.9447 -> 0.9447 | No clear effect |
Closest papers by meaning
- Diffusion4D: Fast Spatial-temporal Consistent 4D generation via Video Diffusion Models (NeurIPS 2024 · similarity 0.92)
- 4Diffusion: Multi-view Video Diffusion Model for 4D Generation (NeurIPS 2024 · similarity 0.92)
- Motion4D: Learning 3D-Consistent Motion and Semantics for 4D Scene Understanding (NeurIPS 2025 · similarity 0.91)
- Track4Gen: Teaching Video Diffusion Models to Track Points Improves Video Generation (CVPR 2025 · similarity 0.90)
- Animate3D: Animating Any 3D Model with Multi-view Video Diffusion (NeurIPS 2024 · similarity 0.90)
- TrackCraft3R: Repurposing Video Diffusion Transformers for Dense 3D Tracking (NeurIPS 2026 · similarity 0.90)
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.
