Research map / Point clouds / Shape and surface completion
ConfCtrl: Enabling Precise Camera Control in Video Diffusion via Confidence-Aware Interpolation
Liudi Yang, George Eskandar, Fengyi Shen, Mohammad Altillawi, Yang Bai, Chi zhang, Ziyuan Liu, Abhinav Valada
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 confidence-weighted noise initialization (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | PSNR: 15.44 -> 15.54 | Improved |
| add confidence map instead of uniform weight 1.0 (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | PSNR: 14.75 -> 15.54 | Improved |
| add update submodule in Kalman DiT block (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | PSNR: 14.65 -> 15.54 | Improved |
| add update submodule and confidence-aware noise init together (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | Translation error E_t: 0.213 -> 0.143 | Improved |
| add gradient regulation loss (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | PSNR: 15.49 -> 15.54 | Improved |
| add projected point cloud cross-attention to the prediction submodule (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | PSNR: 14.90 -> 15.54 | Improved |
| use cross-attention rather than simple addition of pose and projection features (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | PSNR: 15.06 -> 15.54 | Improved |
| initialize from video interpolation model instead of camera-guided video model (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | PSNR: 13.45 -> 15.54 | Improved |
| deepen the design to 15 layers instead of using Kalman DiT blocks (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | PSNR: 15.54 -> 14.56 | Hurt |
| replace VGGT with Pi3 as 3D foundation model (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | PSNR: 15.54 -> 15.56 | No clear effect |
| replace VGGT with Dust3R as 3D foundation model (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | PSNR: 15.54 -> 14.99 | Hurt |
| replace Kalman DiT block with MMDiT block (ablation) | ConfCtrl · CO3D-Hydrant (ablation) | PSNR: 15.54 -> 15.01 | Hurt |
Closest papers by meaning
- ViVid-1-to-3: Novel View Synthesis with Video Diffusion Models (CVPR 2024 · similarity 0.88)
- AC3D: Analyzing and Improving 3D Camera Control in Video Diffusion Transformers (CVPR 2025 · similarity 0.88)
- NAMVIS: Next-Scale Autoregressive Multi-View Image Synthesis (NeurIPS 2026 · similarity 0.88)
- Controlling Space and Time with Diffusion Models (ICLR 2025 · similarity 0.88)
- UMAMI: Unifying Masked Autoregressive Models and Deterministic Rendering for View Synthesis (NeurIPS 2025 · similarity 0.87)
- Dynamic View Synthesis as an Inverse Problem (NeurIPS 2025 · similarity 0.87)
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.
