Research map / 3D vision / 3D generation
MatryODShka: Real-time 6DoF Video View Synthesis using Multi-Sphere Images
Benjamin Attal, Selena Ling, Aaron Gokaslan, Christian Richardt, James Tompkin
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 CoordConv (ablation) | Ours (real-time) · Replica ODS test set | mE-LPIPS: 2.43 -> 2.29 | No clear effect |
| add E-LPIPS perceptual loss instead of L2 (ablation) | Ours (real-time) · Replica ODS test set | mE-LPIPS: 2.88 -> 2.29 | Improved |
| add E-LPIPS perceptual loss instead of L2 (ablation) | Ours (real-time) · Replica ODS test set | PSNR: 29.82 -> 29.10 | Hurt |
| use 64 MSI layers instead of 32 (ablation) | Ours (real-time) · Replica ODS test set | mE-LPIPS: 2.29 -> 2.45 | No clear effect |
| use graph convolutional network instead of convolutions (ablation) | Ours (real-time) · Replica ODS test set | mE-LPIPS: 2.29 -> 3.06 | Hurt |
| use graph convolutional network instead of convolutions (ablation) | Ours (real-time) · Replica ODS test set | PSNR: 29.10 -> 30.03 | Improved |
| add background layer (ablation) | Ours (real-time) · Replica ODS test set | mE-LPIPS: 2.29 -> 2.04 | Improved |
| add background layer (ablation) | Ours (real-time) · Replica ODS test set | PSNR: 29.10 -> 30.03 | Improved |
| train double plane sweep on RealEstate10K instead of Replica | Double plane sweep · Replica ODS test set | mE-LPIPS: 7.46 -> 4.28 | Improved |
| increase transform-inverse regularization range from none to x5.0 (ablation) | Ours · Replica temporal consistency | f2f-depth: 3.27 -> 1.04 | Improved |
| increase transform-inverse regularization range from none to x5.0 (ablation) | Ours · Replica temporal consistency | PSNR: 29.82 -> 29.20 | Hurt |
| add transform-inverse regularization at x0.5 (ablation) | Ours · Replica temporal consistency | mE-LPIPS: 2.88 -> 2.37 | Improved |
Closest papers by meaning
- Imagine360: Immersive 360 Video Generation from Perspective Anchor (NeurIPS 2025 · similarity 0.86)
- Balanced Spherical Grid for Egocentric View Synthesis (CVPR 2023 · similarity 0.86)
- Novel View Synthesis of Dynamic Scenes With Globally Coherent Depths From a Monocular Camera (CVPR 2020 · similarity 0.86)
- Geometry-guided Online 3D Video Synthesis with Multi-View Temporal Consistency (CVPR 2025 · similarity 0.85)
- Depth Anywhere: Enhancing 360 Monocular Depth Estimation via Perspective Distillation and Unlabeled Data Augmentation (NeurIPS 2024 · similarity 0.85)
- MultiDiff: Consistent Novel View Synthesis from a Single Image (CVPR 2024 · similarity 0.84)
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.
