Research map / Video / Video generation
Geometric Reciprocity: Unlocking Self-Supervision for Stereoscopic Video Generation
Jingyi Lu, Kai Han
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 |
|---|---|---|---|
| fine-tune on GRT data | LaMa · DAVIS-GRT | PSNR: 31.75 -> 35.52 | Improved |
| fine-tune on GRT data | SD1.5 Inpainting · DAVIS-GRT | PSNR: 27.38 -> 30.70 | Improved |
| fine-tune on GRT data | SD2.1 Inpainting · DAVIS-GRT | PSNR: 27.37 -> 28.12 | Improved |
| fine-tune on GRT data | SDXL Inpainting · DAVIS-GRT | PSNR: 23.40 -> 29.45 | Improved |
| fine-tune on GRT data | ProPainter · DAVIS-GRT | PSNR: 31.03 -> 34.06 | Improved |
| fine-tune on GRT data | StereoCrafter · DAVIS-GRT | PSNR: 28.95 -> 30.85 | Improved |
| add depth-occlusion term to boundary-violation-only GRT mask (ablation) | LaMa on ImageNet-GRT · DAVIS-GRT | PSNR: 32.89 -> 35.52 | Improved |
| add boundary-violation term to depth-occlusion-only GRT mask (ablation) | LaMa on ImageNet-GRT · DAVIS-GRT | PSNR: 32.54 -> 35.52 | Improved |
| use GRT analytic masks instead of naive cycle consistency | LaMa · DAVIS-GRT | PSNR: 32.14 -> 33.10 | Improved |
| use GRT instead of naive cycle consistency (training speed) | LaMa · DAVIS-GRT | steps/s: 0.6 -> 1.9 | Improved |
| use stronger depth estimator (V1-Small to V2-Large) (ablation) | Ours-Video · Inria 3DMovie | MEt3R: 0.1058 -> 0.0973 | Improved |
| increase training data from 0% to 100% Kinetics-GRT (ablation) | ProPainter · DAVIS-GRT | PSNR: 31.03 -> 34.06 | Improved |
Closest papers by meaning
- SpatialDreamer: Self-supervised Stereo Video Synthesis from Monocular Input (CVPR 2025 · similarity 0.89)
- Reconstruct, Inpaint, Test-Time Finetune: Dynamic Novel-view Synthesis from Monocular Videos (NeurIPS 2025 · similarity 0.88)
- Reversing the cycle: self-supervised deep stereo through enhanced monocular distillation (ECCV 2020 · similarity 0.87)
- Self-Supervised Learning With Geometric Constraints in Monocular Video: Connecting Flow, Depth, and Camera (ICCV 2019 · similarity 0.87)
- StereoWorld: Geometry-Aware Monocular-to-Stereo Video Generation (CVPR 2026 · similarity 0.87)
- SVG: 3D Stereoscopic Video Generation via Denoising Frame Matrix (ICLR 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.
