Research map / Label-efficient and robust learning / Self-supervised and contrastive
Semi-Supervised Learning of Optical Flow by Flow Supervisor
Woobin Im, Sebin Lee, Sung-Eui Yoon
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 parameter separation (separate supervisor network) (ablation) | RAFT + FS · Sintel/KITTI | EPE: diverged -> 1.30 / 2.46 / 3.35 / 11.12 | Improved |
| pass student outputs with residual connection to the supervisor (ablation) | RAFT + FS · KITTI (train) | EPE: 6.68 -> 3.35 | Improved |
| pass student outputs with residual connection (ablation) | RAFT + FS · Sintel (train) | EPE: 1.30 -> 1.30 | No clear effect |
| add residual connection on top of passing student outputs (ablation) | RAFT + FS · KITTI (train) | EPE: 6.16 -> 3.35 | Improved |
| share encoder between student and supervisor (ablation) | RAFT + FS · KITTI (train) | EPE: 3.80 -> 3.35 | Improved |
| share encoder between student and supervisor (ablation) | RAFT + FS · Sintel Final (train) | EPE: 2.58 -> 2.46 | Improved |
| add unsupervised teacher loss (+TU) to supervised teacher loss (ablation) | RAFT + FS · KITTI (train) | EPE: 4.69 -> 3.35 | Improved |
| add unsupervised teacher loss (+TU) (ablation) | RAFT + FS · Sintel Final (train) | EPE: 2.46 -> 2.80 | Hurt |
| use clean teacher input instead of augmented (ablation) | RAFT + FS · KITTI (train) | EPE: 4.17 -> 3.35 | Improved |
| stop gradient on teacher encoder (ablation) | RAFT + FS · Sintel Final (train) | EPE: 2.58 -> 2.46 | Improved |
| flow supervisor vs naive self-supervision (Sup + Self) | RAFT · Sintel/KITTI | EPE: diverged -> 1.30 / 2.46 / 3.35 | Improved |
| flow supervisor vs fixed teacher | RAFT · KITTI (train) | EPE: 4.91 -> 3.35 | Improved |
Closest papers by meaning
- SelFlow: Self-Supervised Learning of Optical Flow (CVPR 2019 · similarity 0.91)
- Semi-Supervised Learning for Optical Flow with Generative Adversarial Networks (NeurIPS 2017 · similarity 0.90)
- Optical Flow Training under Limited Label Budget via Active Learning (ECCV 2022 · similarity 0.90)
- Learning by Distillation: A Self-Supervised Learning Framework for Optical Flow Estimation (TPAMI 2021 · similarity 0.89)
- Learning by Analogy: Reliable Supervision From Transformations for Unsupervised Optical Flow Estimation (CVPR 2020 · similarity 0.88)
- Recurrent Partial Kernel Network for Efficient Optical Flow Estimation (AAAI 2024 · 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.
