Research map / Point clouds / Point cloud classification
EPiC: Efficient Video Camera Control Learning with Precise Anchor-Video Guidance
Zun Wang, Jaemin Cho, Jialu Li, Han Lin, Jaehong Yoon, Yue Zhang, Mohit Bansal
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 |
|---|---|---|---|
| use masking-based anchors instead of point-cloud-rendered anchors for training (ablation) | EPiC Anchor-ControlNet · RE10K anchor-type ablation | RotErr: 0.60 -> 0.40 | Improved |
| use masking-based anchors instead of point-cloud anchors (ablation) | EPiC Anchor-ControlNet · RE10K anchor-type ablation | TransErr: 1.07 -> 0.86 | Improved |
| use masking-based anchors instead of point-cloud anchors (ablation) | EPiC Anchor-ControlNet · RE10K anchor-type ablation | anchor PSNR: 16.01 -> 40.12 | Improved |
| mix 50% point-cloud anchors with 50% masking anchors (ablation) | EPiC Anchor-ControlNet · RE10K anchor-type ablation | CamMC: 1.45 -> 1.29 (vs 1.17 for fully masking) | Mixed |
| inject synthetic flying-pixel artifacts into training anchors (ablation) | EPiC · qualitative ablation | visual artifacts | Improved |
| add visibility-aware output masking to ControlNet (ablation) | EPiC Anchor-ControlNet · qualitative ablation | visual quality in invisible regions | Improved |
| mask dynamic objects out of the point cloud at inference (ablation) | EPiC I2V inference · qualitative ablation | dynamic object motion | Improved |
| use EPiC (5K videos, 500 steps, <30M params) instead of prior camera-control baselines | I2V camera control · RE10K | TransErr: 0.99 -> 0.86 (Gen3C vs EPiC) | Improved |
| use EPiC instead of baselines on visual quality | I2V camera control · RE10K | VBench total score: 82.27 -> 82.63 (Gen3C vs EPiC) | No clear effect |
| apply EPiC zero-shot to V2V trained only on I2V data | EPiC · Kubric-4D | PSNR: 19.72 (GCD) vs 19.65 (EPiC) | No clear effect |
Closest papers by meaning
- Training-free Camera Control for Video Generation (ICLR 2025 · similarity 0.88)
- AC3D: Analyzing and Improving 3D Camera Control in Video Diffusion Transformers (CVPR 2025 · similarity 0.87)
- Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance (NeurIPS 2025 · similarity 0.87)
- $h$-control: Training-Free Camera Control via Block-Conditional Gibbs Refinement (NeurIPS 2026 · similarity 0.87)
- VD3D: Taming Large Video Diffusion Transformers for 3D Camera Control (ICLR 2025 · similarity 0.86)
- GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control (CVPR 2025 · 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.
