Research map / Video / Video generation
MotionEditGS: Editing Motion and Appearance of 4D Scenes from Monocular Video via Semantically Anchored Gaussians
Daehyun We, Youngho Yoon, Jiyong Boo, KUK-JIN YOON
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 semantically anchored editing (SAE) over RGB-only (ablation) | MotionEditGS · reconstruction of edited video, look left | PSNR: 30.73 -> 38.15 | Improved |
| add SAE over RGB-only (ablation) | MotionEditGS · reconstruction of edited video, look left | LPIPS: 0.164 -> 0.084 | Improved |
| add gradient-based feature lifting to SAE (ablation) | MotionEditGS · reconstruction of edited video, look left | PSNR: 38.15 -> 38.36 | No clear effect |
| add feature lifting to SAE (ablation) | MotionEditGS · reconstruction of edited video, blue cap | PSNR: 17.32 -> 37.80 | Improved |
| add feature lifting to SAE (ablation) | MotionEditGS · reconstruction of edited video, blue cap | LPIPS: 0.423 -> 0.074 | Improved |
| add SAE alone (no lifting) over RGB-only (ablation) | MotionEditGS · reconstruction of edited video, blue cap | PSNR: 34.88 -> 17.32 | Hurt |
| add semantic feature supervision vs RGB-only baseline | MotionEditGS · mochi-high-five | LOVE-Correspondence novel view: 0.547 -> 0.609 | Improved |
| add semantic feature supervision vs RGB-only baseline | MotionEditGS · mochi-high-five | VBench consistency novel view: 0.927 -> 0.947 | Improved |
| add semantic feature supervision vs RGB-only baseline | MotionEditGS · paper-windmill | VBench consistency novel view: 0.899 -> 0.898 | No clear effect |
| add semantic feature supervision vs RGB-only baseline | MotionEditGS · mochi-high-five | CLIP train view: 33.56 -> 32.78 | Hurt |
| add semantic feature supervision vs RGB-only baseline | MotionEditGS · pillow | CLIP novel view: 22.17 -> 23.65 | Improved |
| use MotionEditGS instead of CTRL-D | MotionEditGS · DyCheck-iPhone | LOVE-Correspondence novel view: 0.467 -> 0.530 | Improved |
Closest papers by meaning
- Instruct 4D-to-4D: Editing 4D Scenes as Pseudo-3D Scenes Using 2D Diffusion (CVPR 2024 · similarity 0.91)
- Efficient Dynamic Scene Editing via 4D Gaussian-based Static-Dynamic Separation (CVPR 2025 · similarity 0.90)
- Motion4D: Learning 3D-Consistent Motion and Semantics for 4D Scene Understanding (NeurIPS 2025 · similarity 0.90)
- CTRL-D: Controllable Dynamic 3D Scene Editing with Personalized 2D Diffusion (CVPR 2025 · similarity 0.89)
- EditSplat: Multi-View Fusion and Attention-Guided Optimization for View-Consistent 3D Scene Editing with 3D Gaussian Splatting (CVPR 2025 · similarity 0.89)
- Align Your Gaussians: Text-to-4D with Dynamic 3D Gaussians and Composed Diffusion Models (CVPR 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.
