Research map / 3D vision / Depth and stereo
Self-Supervised Multi-Frame Monocular Scene Flow
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace two-frame estimation with three-frame estimation and forward-warped ConvLSTM state propagation (ablation) | Advanced two-frame baseline · KITTI Scene Flow Training | SF-all: 46.70 -> 42.37 | Helped |
| Add the occlusion-aware census loss (ablation) | Multi-Mono-SF · KITTI Scene Flow Training | SF-all: 42.37 -> 39.82 | Helped |
| Use three frames and bidirectional cost volumes without ConvLSTM (ablation) | Advanced two-frame baseline · KITTI Scene Flow Training | SF-all: 46.70 -> 46.15 | Helped |
| Add ConvLSTM state propagation without warping (ablation) | Three-frame estimation · KITTI Scene Flow Training | SF-all: 46.15 -> 45.43 | Helped |
| Use backward rather than forward warping for ConvLSTM state propagation (ablation) | Three-frame ConvLSTM · KITTI Scene Flow Training | SF-all: 45.43 -> 46.90 | Hurt |
| Replace standard census with occlusion-aware census (ablation) | Multi-frame model · KITTI Scene Flow Training | SF-all: 41.71 -> 39.82 | Helped |
| Replace brightness difference + SSIM with occlusion-aware census (ablation) | Multi-frame model · KITTI Scene Flow Training | SF-all: 42.37 -> 39.82 | Helped |
| Remove the context network (ablation) | Self-Mono-SF on the multi-frame train split · KITTI Scene Flow Training | SF-all: 60.88 -> 50.49 | Helped |
| Add cost volume normalization and use one fewer pyramid level (ablation) | No-context Self-Mono-SF · KITTI Scene Flow Training | SF-all: 50.49 -> 46.70 | Helped |
| Split the decoder at the second-to-last layer instead of retaining one joint decoder (ablation) | Self-Mono-SF without context network · KITTI Scene Flow Training | SF-all: 48.97 -> 47.53 | Helped |
| Use one fewer pyramid level (ablation) | No-context baseline with cost volume normalization · KITTI Scene Flow Training | SF-all: 48.29 -> 47.77 | Helped |
| Add data distillation (ablation) | No-context baseline with cost volume normalization · KITTI Scene Flow Training | SF-all: 48.29 -> 48.11 | Helped |
Closest papers by meaning
- Self-Supervised Monocular Scene Flow Estimation (CVPR 2020 · similarity 0.94)
- Zero-Shot Monocular Scene Flow Estimation in the Wild (CVPR 2025 · similarity 0.88)
- DeltaFlow: An Efficient Multi-frame Scene Flow Estimation Method (NeurIPS 2025 · similarity 0.87)
- Masked Scene Modeling: Narrowing the Gap Between Supervised and Self-Supervised Learning in 3D Scene Understanding (CVPR 2025 · similarity 0.86)
- Mining Supervision for Dynamic Regions in Self-Supervised Monocular Depth Estimation (CVPR 2024 · similarity 0.85)
- Semantic Flow: Learning Semantic Fields of Dynamic Scenes from Monocular Videos (ICLR 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.