Research map / 3D vision / Depth and stereo
Estimating 2D Camera Motion with Hybrid Motion Basis
Haipeng Li, Tianhao Zhou, Zhanglei Yang, Yi Wu, Yan Chen, Zijing Mao, Shen Cheng, Bing Zeng, Shuaicheng Liu
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 |
|---|---|---|---|
| increase motion bases from 8 to 12 physical bases (ablation) | CamFlow · GHOF-Cam | EPE: 1.45 -> 1.23 | Improved |
| add stochastic bases to reach 24 bases (ablation) | CamFlow · GHOF-Cam | EPE: 1.23 -> 1.10 | Improved |
| add stochastic bases on GHOF zero-shot (ablation) | CamFlow · GHOF | PME: 1.54 -> 1.23 | Improved |
| increase bases from 24 to 200 (ablation) | CamFlow · GHOF-Cam | EPE: 1.10 -> 1.07 | No clear effect |
| increase bases from 24 to 200 (inference time) (ablation) | CamFlow · Inference time | ms: 79.63ms -> 99.28ms | Hurt |
| add photometric loss to motion-supervision loss (ablation) | CamFlow · GHOF-Cam | EPE: 2.13 -> 1.10 | Improved |
| add motion-supervision loss to photometric loss (ablation) | CamFlow · GHOF-Cam | EPE: 1.42 -> 1.10 | Improved |
| add photometric loss to motion-supervision loss (ablation) | CamFlow · CAHomo | PME: 0.41 -> 0.33 | Improved |
| full training instead of early stopping (ablation) | CamFlow · CAHomo | PME: 0.33 -> 0.32 | No clear effect |
| use hybrid motion bases instead of 8 linear homography bases | BasesHomo · GHOF-Cam | EPE: 1.74 -> 1.10 | Improved |
| use CamFlow instead of HomoGAN | HomoGAN · CAHomo | PME: 0.39 -> 0.32 | Improved |
| use CamFlow instead of RealSH | RealSH · GHOF low-light | PME: 4.42 -> 2.69 | Improved |
Closest papers by meaning
- CamLiFlow: Bidirectional Camera-LiDAR Fusion for Joint Optical Flow and Scene Flow Estimation (CVPR 2022 · similarity 0.86)
- Optical Flow in Mostly Rigid Scenes (CVPR 2017 · similarity 0.85)
- VideoFlow: Exploiting Temporal Cues for Multi-frame Optical Flow Estimation (ICCV 2023 · similarity 0.85)
- DeltaFlow: An Efficient Multi-frame Scene Flow Estimation Method (NeurIPS 2025 · similarity 0.85)
- ZeroFlow: Scalable Scene Flow via Distillation (ICLR 2024 · similarity 0.84)
- Zero-Shot Monocular Scene Flow Estimation in the Wild (CVPR 2025 · 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.
