Research map / 3D vision / Depth and stereo
UFO-4D: Unposed Feedforward 4D reconstruction from Two Images
Junhwa Hur, Charles Herrmann, Songyou Peng, Philipp Henzler, Zeyu Ma, Todd Zickler, Deqing 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 photometric-loss gradient to Gaussian center and velocity (full model vs w/o image gradient) (ablation) | UFO-4D · Stereo4D (256x256 ablation) | Point EPE: 0.903 -> 0.827 | Improved |
| add rendering losses on point and motion (full vs w/o motion and point rendering) (ablation) | UFO-4D · Stereo4D (256x256 ablation) | Point EPE: 0.911 -> 0.827 | Improved |
| add rendering losses on point and motion (ablation) | UFO-4D · Stereo4D (256x256 ablation) | PSNR: 20.675 -> 23.929 | Improved |
| add all rendering losses (point, motion, image) versus none (ablation) | UFO-4D · Stereo4D (256x256 ablation) | Point EPE: 0.886 -> 0.827 | Improved |
| add image gradient to Gaussian parameters (ablation) | UFO-4D · Stereo4D (256x256 ablation) | PSNR: 24.268 -> 23.929 | Hurt |
| use dynamic 3DGS output instead of per-pixel point and motion (ablation) | UFO-4D · KITTI | Point EPE: 4.567 -> 3.553 | Improved |
| use dynamic 3DGS output instead of per-pixel point and motion (ablation) | UFO-4D · Stereo4D | Point EPE: 0.809 -> 0.781 | Improved |
| use dynamic 3DGS output instead of per-pixel point and motion (ablation) | UFO-4D · Bonn | Point EPE: 0.196 -> 0.211 | Hurt |
| use feedforward pose head instead of PnP+RANSAC on UFO-4D (ablation) | UFO-4D · Stereo4D | ATE: 0.0120 -> 0.0101 | Improved |
| initialize from MASt3R instead of MonST3R (ablation) | UFO-4D · Stereo4D (small scale) | Point EPE: 0.868 -> 0.781 | Improved |
| initialize from MASt3R instead of MonST3R (ablation) | UFO-4D · Bonn (small scale) | Point EPE: 0.155 -> 0.211 | Hurt |
| train on Stereo4D only instead of mixing ST/PO/VK (ablation) | UFO-4D · Stereo4D (dataset mixture) | PSNR: 21.918 -> 23.929 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- MonST3R — MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion (ICLR 2025) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Flux4D: Flow-based Unsupervised 4D Reconstruction (NeurIPS 2025 · similarity 0.88)
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.87)
- Instant4D: 4D Gaussian Splatting in Minutes (NeurIPS 2025 · similarity 0.87)
- Real-time Photorealistic Dynamic Scene Representation and Rendering with 4D Gaussian Splatting (ICLR 2024 · similarity 0.87)
- 4D-LRM: Large Space-Time Reconstruction Model From and To Any View at Any Time (NeurIPS 2025 · similarity 0.86)
- Stereo4D: Learning How Things Move in 3D from Internet Stereo Videos (CVPR 2025 · similarity 0.86)
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.
