Research map / Video / Video segmentation and tracking
SMURF: Self-Teaching Multi-Frame Unsupervised RAFT With Full-Image Warping
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 |
|---|---|---|---|
| Train with SMURF rather than UFlow | Unsupervised optical flow · Sintel Clean test | EPE: 40% | Helped |
| Train with SMURF rather than UFlow | Unsupervised optical flow · Sintel Final test | EPE: 36% | Helped |
| Train with SMURF rather than UFlow | Unsupervised optical flow · KITTI 2015 test | ER: 39% | Helped |
| Replace UFlow's PWC model with RAFT without changing the unsupervised training method | UFlow · Sintel Clean train | EPE: 3.01 -> 3.36 | Hurt |
| Add sequence loss (ablation) | SMURF with RAFT · Sintel Clean train | EPE: 2.92 -> 2.66 | Helped |
| Add heavy augmentation to sequence loss (ablation) | SMURF with RAFT · Sintel Clean train | EPE: 2.66 -> 2.38 | Helped |
| Add full-image warping to sequence loss and heavy augmentation (ablation) | SMURF with RAFT · KITTI-15 train | EPE: 3.64 -> 2.45 | Helped |
| Add multi-frame self-supervision to sequence loss, heavy augmentation, and full-image warping (ablation) | SMURF with RAFT · KITTI-15 train | EPE: 2.45 -> 2.01 | Helped |
| Add multi-frame self-supervision to sequence loss, heavy augmentation, and full-image warping (ablation) | SMURF with RAFT · Sintel Clean train | EPE: 2.15 -> 1.99 | Helped |
| Remove self-supervision (ablation) | SMURF with RAFT · Sintel Clean train | EPE: 2.15 -> 3.88 | Hurt |
| Use final rather than intermediate predictions to generate self-supervision labels (ablation) | SMURF with RAFT · Sintel Clean train | EPE: 2.51 -> 2.15 | Helped |
| Remove forward-backward consistency masking from self-supervision (ablation) | SMURF with RAFT · Sintel Clean train | EPE: 2.53 -> 2.15 | Helped |
Closest papers by meaning
- Learning by Analogy: Reliable Supervision From Transformations for Unsupervised Optical Flow Estimation (CVPR 2020 · similarity 0.89)
- UnSAMFlow: Unsupervised Optical Flow Guided by Segment Anything Model (CVPR 2024 · similarity 0.88)
- OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation (CVPR 2024 · similarity 0.85)
- Taming generative video models for zero-shot optical flow extraction (NeurIPS 2025 · similarity 0.85)
- M2Flow: A Motion Information Fusion Framework for Enhanced Unsupervised Optical Flow Estimation in Autonomous Driving (AAAI 2025 · similarity 0.84)
- Recurrent Partial Kernel Network for Efficient Optical Flow Estimation (AAAI 2024 · 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.