Research map / 3D vision / Depth and stereo
Consistency Guided Scene Flow Estimation
Yuhua Chen, Luc Van Gool, Cordelia Schmid, Cristian Sminchisescu
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 consistency guided refinement network | feedforward module · KITTI 2015 scene flow | SF outliers (%): 43.55 -> 35.40 | Improved |
| add consistency guided refinement network | feedforward module · FlyingThings3D test | flow EPE: 7.02 -> 5.98 | Improved |
| add self-supervised finetuning on KITTI after synthetic pretraining | CGSF · KITTI 2015 scene flow | SF outliers (%): 35.40 -> 28.14 | Improved |
| add refinement without consistency inputs (prediction only) (ablation) | feedforward · KITTI 2015 SF outliers ablation (trained on FlyingThings3D) | SF outliers (%): 43.01 -> 42.43 | Improved |
| add stereo photometric consistency L_pd (ablation) | refinement network · KITTI 2015 SF outliers ablation (trained on FlyingThings3D) | SF outliers (%): 42.43 -> 39.80 | Improved |
| add flow photometric consistency L_pf (ablation) | refinement network · KITTI 2015 SF outliers ablation (trained on FlyingThings3D) | SF outliers (%): 42.43 -> 39.64 | Improved |
| add disparity-flow consistency L_df (ablation) | refinement network · KITTI 2015 SF outliers ablation (trained on FlyingThings3D) | SF outliers (%): 42.43 -> 39.59 | Improved |
| add L_df to L_pd and L_pf (ablation) | refinement network · KITTI 2015 SF outliers ablation (trained on FlyingThings3D) | SF outliers (%): 38.23 -> 35.40 | Improved |
| add loss as input to refinement network (ablation) | refinement network · KITTI 2015 SF outliers ablation (trained on FlyingThings3D) | SF outliers (%): 40.47 -> 35.40 | Improved |
| add gradient as input to refinement network (ablation) | refinement network · KITTI 2015 SF outliers ablation (trained on FlyingThings3D) | SF outliers (%): 36.31 -> 35.40 | Improved |
| learned refinement instead of output finetuning (ablation) | refinement strategy · KITTI 2015 online refinement strategies | SF outliers (%): 39.95 -> 35.40 | Improved |
| learned refinement instead of parameter finetuning (ablation) | refinement strategy · KITTI 2015 online refinement strategies | SF outliers (%): 35.49 -> 35.40 | No clear effect |
Closest papers by meaning
- Imposing Consistency for Optical Flow Estimation (CVPR 2022 · similarity 0.90)
- Triangular Consistency as a Universal Constraint for Learning Optical Flow (ECCV 2026 · similarity 0.88)
- DifFlow3D: Toward Robust Uncertainty-Aware Scene Flow Estimation with Iterative Diffusion-Based Refinement (CVPR 2024 · similarity 0.87)
- DeltaFlow: An Efficient Multi-frame Scene Flow Estimation Method (NeurIPS 2025 · similarity 0.87)
- CH3Depth: Efficient and Flexible Depth Foundation Model with Flow Matching (CVPR 2025 · similarity 0.86)
- MESSENGER: Memory-Enhanced Sequential Scene Flow Estimation via Autoregressive Next-Frame Forecasting (NeurIPS 2026 · 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.
