Research map / Video / Video generation
AdaViewPlanner: Adapting Video Diffusion Models for Viewpoint Planning in 4D Scenes
Yu Li, Menghan Xia, Gongye Liu, Jianhong Bai, Xintao Wang, Conglang Zhang, Yuxuan Lin, Ruihang Chu, Pengfei Wan, Yujiu Yang
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 random camera-pose guidance during Stage I video generation training (ablation) | AdaViewPlanner Stage I · General-domain 4D human motion control | WA-MPJPE (mm): 127.68 -> 103.92 | Improved |
| add random camera-pose guidance during Stage I video generation training (ablation) | AdaViewPlanner Stage I · TikTok (dance domain) | WA-MPJPE (mm): 72.60 -> 71.65 | Improved |
| add random camera-pose guidance during Stage I video generation training (ablation) | AdaViewPlanner Stage I · General domain | PA-MPJPE (mm): 33.80 -> 35.70 | Hurt |
| add 3D RoPE for human motion tokens (ablation) | AdaViewPlanner Stage I · General-domain 4D human motion control | WA-MPJPE (mm): 122.13 -> 103.92 | Improved |
| add 3D RoPE for human motion tokens (ablation) | AdaViewPlanner Stage I · TikTok (dance domain) | WA-MPJPE (mm): 82.59 -> 71.65 | Improved |
| train Stage I to full convergence instead of an early checkpoint (ablation) | AdaViewPlanner · camera extraction reprojection | reprojection MSE: 0.181 -> 0.158 | Improved |
| add human motion conditioning to Stage II camera extraction (ablation) | AdaViewPlanner Stage II · camera extraction reprojection | reprojection IoU: 0.226 -> 0.338 | Improved |
| predict absolute camera poses with motion branch instead of relative poses (ablation) | AdaViewPlanner Stage II · camera extraction reprojection | reprojection MSE: 0.167 -> 0.158 | Improved |
| predict absolute camera poses with motion branch instead of relative poses (ablation) | AdaViewPlanner Stage II · Human Missing Rate | HMR: 0.039 -> 0.018 | Improved |
| use video-model-based two-stage planning instead of re-implemented DanceCamera3D | AdaViewPlanner · E.T. testset | Text-Camera Consistency: 0.975 -> 1.125 | Improved |
| use video-model-based two-stage planning instead of E.T. | AdaViewPlanner · E.T. testset | user preference (%): 23.81 -> 61.90 | Improved |
| video-model-based planning vs E.T. on camera translation jerk | AdaViewPlanner · E.T. testset | Jerk_t: 0.001 -> 0.007 | Hurt |
Closest papers by meaning
- Diffusion$^2$: Dynamic 3D Content Generation via Score Composition of Video and Multi-view Diffusion Models (ICLR 2025 · similarity 0.89)
- CAT4D: Create Anything in 4D with Multi-View Video Diffusion Models (CVPR 2025 · similarity 0.89)
- 4Diffusion: Multi-view Video Diffusion Model for 4D Generation (NeurIPS 2024 · similarity 0.89)
- SV4D: Dynamic 3D Content Generation with Multi-Frame and Multi-View Consistency (ICLR 2025 · similarity 0.89)
- ViVid-1-to-3: Novel View Synthesis with Video Diffusion Models (CVPR 2024 · similarity 0.89)
- Vivid-ZOO: Multi-View Video Generation with Diffusion Model (NeurIPS 2024 · similarity 0.88)
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.
