Research map / 3D vision / Gaussian splatting
Learning Video Dynamics with Predictive Differentiable Rendering
Yujin Tang, Tian Zhou, Xin Lin, Cheng Tan, Yifan Hu, Rong Jin, SouYoung Jin, Liang Sun
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 PredGS Gaussian adapter with L1+SSIM loss | TAU · TaxiBJ (4->4 frames) | MSE: 0.3108 -> 0.2807 | Improved |
| add PredGS Gaussian adapter with L1+SSIM loss | TAU · WeatherBench T2m (12->12 frames) | MSE: 1.162 -> 1.071 | Improved |
| add PredGS Gaussian adapter with L1+SSIM loss | TAU · KTH (10->20 frames) | LPIPS: 0.22856 -> 0.16237 | Improved |
| add PredGS Gaussian adapter with L1+SSIM loss | TAU · Human3.6M (4->4 frames) | LPIPS: 0.02783 -> 0.02192 | Improved |
| add PredGS Gaussian adapter with L1+SSIM loss | TAU · KTH (10->20 frames) | PSNR: 27.10 -> 26.98 | Hurt |
| replace L2+reg with L1+SSIM loss (ablation) | TAU · TaxiBJ (4->4 frames) | MSE: 0.3108 -> 0.3088 | Improved |
| add PredGS adapter while keeping L2+reg loss (ablation) | TAU · TaxiBJ (4->4 frames) | MSE: 0.3108 -> 0.3400 | Hurt |
| replace L2+reg with L1+SSIM loss (ablation) | PDR · TaxiBJ (4->4 frames) | MSE: 0.3400 -> 0.2807 | Improved |
| add uniform grid Gaussian coordinate initialization (ablation) | PDR · TaxiBJ (4->4 frames) | MSE: 0.2893 -> 0.2807 | Improved |
| fuse by channel concat plus conv instead of addition (ablation) | PDR · TaxiBJ (4->4 frames) | MSE: 0.2998 -> 0.2807 | Improved |
| fuse pixel prediction with Gaussian branch instead of using only Gaussian branch (ablation) | PDR · TaxiBJ (4->4 frames) | MSE: 0.3501 -> 0.2807 | Improved |
| use balanced loss weight lambda 0.5 instead of 1 (L1 only) (ablation) | PDR · TaxiBJ (4->4 frames) | MSE: 0.2907 -> 0.2807 | Improved |
Closest papers by meaning
- 3DGS-Enhancer: Enhancing Unbounded 3D Gaussian Splatting with View-consistent 2D Diffusion Priors (NeurIPS 2024 · similarity 0.87)
- HiCoM: Hierarchical Coherent Motion for Dynamic Streamable Scenes with 3D Gaussian Splatting (NeurIPS 2024 · similarity 0.87)
- World from Motion: Generative Dynamic Gaussian Reconstruction from Monocular Video (NeurIPS 2026 · similarity 0.87)
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.87)
- Diffusion Renderer: Neural Inverse and Forward Rendering with Video Diffusion Models (CVPR 2025 · similarity 0.87)
- Real-time Photorealistic Dynamic Scene Representation and Rendering with 4D Gaussian Splatting (ICLR 2024 · 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.
