Research map / 3D vision / Depth and stereo
Differential Scene Flow from Light Field Gradients
Sizhuo Ma, Brandon M. Smith, Mohit Gupta
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 |
|---|---|---|---|
| estimate Z-motion directly with structure-aware global ray flow instead of via depth differences (PD-Flow, OLFW) | SAG ray flow · Translation stage playing cards (Lytro Illum) | Z-motion MAE | Improved |
| add structure-aware global regularization to local LK and global HS ray flow (ablation) | ray flow · Translation stage controlled experiments | motion accuracy | Improved |
| increase amount of X-motion beyond 3.5mm (ablation) | SAG ray flow · Single textured plane, X-motion | RMSE: error increases rapidly when the motion is larger than 3.5mm | Hurt |
| increase effective aperture size from 1.8mm to 3.6mm to 7.2mm (ablation) | SAG ray flow · Simulated single plane, X-motion | mean relative error: 1.8mm, 3.6mm and 7.2mm | Improved |
| increase angular resolution from 7x7 to 13x13 to 25x25 (ablation) | SAG ray flow · Simulated single plane, small motion | mean relative error: 7x7, 13x13 and 25x25 | Improved |
| increase angular resolution for large motion (ablation) | SAG ray flow · Simulated single plane, large motion | mean relative error: 7x7, 13x13 and 25x25 | No clear effect |
| increase surface specularity k from 0 to 0.3 (ablation) | SAG ray flow · Simulated rotating plane, Phong BRDF | error: k from 0 to 0.3 | Hurt |
| increase scene depth (ablation) | SAG ray flow · Simulated plane at varying depth | motion error | Hurt |
| add Gaussian noise to the depth input (ablation) | PD-Flow · Simulated two planes at 0.3m and 0.4m, 2.24mm motion | mean relative error | Hurt |
| add Gaussian noise to the depth input (ablation) | SAG ray flow · Simulated two planes at 0.3m and 0.4m, 2.24mm motion | mean relative error | No clear effect |
| increase amount of rotation (ablation) | SAG ray flow · Simulated rotating plane, Lambertian k=0 | error | Hurt |
Closest papers by meaning
- 3D Scene Flow from 4D Light Field Gradients (ECCV 2018 · similarity 0.99)
- Depth Estimation Using Structured Light Flow -- Analysis of Projected Pattern Flow on an Object's Surface (ICCV 2017 · similarity 0.89)
- SEA-Flow3D: Simplified, Efficient, and Accurate Scene Flow via Spatial Vector Sampling and Multi-scale Refinement (CVPR 2026 · similarity 0.88)
- LiFF: Light Field Features in Scale and Depth (CVPR 2019 · similarity 0.86)
- Ray Space Features for Plenoptic Structure-From-Motion (ICCV 2017 · similarity 0.86)
- iToF-flow-based High Frame Rate Depth Imaging (CVPR 2024 · similarity 0.85)
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.
