Research map / Video / Video segmentation and tracking
MEMFOF: High-Resolution Training for Memory-Efficient Multi-Frame Optical Flow Estimation
Vladislav Bargatin, Egor Chistov, Alexander Yakovenko, Dmitriy Vatolin
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 |
|---|---|---|---|
| train on 2x upsampled full frames instead of native-resolution data (ablation) | MEMFOF · Spring train | EPE: 0.430 -> 0.341 | Improved |
| use full upsampled frames instead of crops (ablation) | MEMFOF · Spring train | EPE: 0.378 -> 0.341 | Improved |
| predict flow at full resolution instead of half-scale inference (native training) (ablation) | MEMFOF · Spring train | EPE: 0.402 -> 0.430 | Mixed |
| bidirectional instead of unidirectional flow (ablation) | MEMFOF · Spring train | EPE: 0.400 -> 0.341 | Improved |
| correlation volume at 1/16 instead of 1/24 resolution (ablation) | MEMFOF (3 frames) · Spring train | 1px outlier rate: 3.480 -> 3.061 | Improved |
| increase update dimension Dc from 128 to 512 (ablation) | MEMFOF (3 frames) · Spring train | 1px outlier rate: 3.560 -> 3.061 | Improved |
| add GMA module (ablation) | MEMFOF (3 frames) · Spring train | 1px outlier rate: 3.151 -> 3.061 | Improved |
| add GMA module (ablation) | MEMFOF (3 frames) · Spring train | EPE: 0.332 -> 0.341 | Hurt |
| use 3 frames instead of 2 frames (ablation) | MEMFOF (1/16) · Spring train | 1px outlier rate: 3.375 -> 3.061 | Improved |
| use 5 frames instead of 3 frames (ablation) | MEMFOF (1/24) · Spring train | 1px outlier rate: 3.480 -> 3.809 | Hurt |
| add inference-time optimizations (late upsampling, feature/correlation reuse, fast correlation) (ablation) | MEMFOF (3 frames) · Inference time | time (ms): 611 -> 472 | Improved |
| increase iterations N from 1 to 8 (ablation) | MEMFOF · Spring train | 1px outlier rate: 3.752 -> 3.061 | Improved |
Later papers that compared with this method
1 later accepted paper reports MEMFOF as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- ARFlow: Auto-regressive Optical Flow Estimation for Arbitrary-Length Videos via Progressive Next-Frame Forecasting (ICLR 2026 · better on 11 of 12 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- StreamFlow — StreamFlow: Streamlined Multi-Frame Optical Flow Estimation for Video Sequences (NeurIPS 2024) · better on 5 of 5 shared results
- MemFlow — MemFlow: Optical Flow Estimation and Prediction with Memory (CVPR 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MemFlow: Optical Flow Estimation and Prediction with Memory (CVPR 2024 · similarity 0.93)
- Removing Cost Volumes from Optical Flow Estimators (ICCV 2025 · similarity 0.90)
- DPFlow: Adaptive Optical Flow Estimation with a Dual-Pyramid Framework (CVPR 2025 · similarity 0.89)
- StreamFlow: Streamlined Multi-Frame Optical Flow Estimation for Video Sequences (NeurIPS 2024 · similarity 0.89)
- Recurrent Partial Kernel Network for Efficient Optical Flow Estimation (AAAI 2024 · similarity 0.87)
- FlowDiffuser: Advancing Optical Flow Estimation with Diffusion Models (CVPR 2024 · similarity 0.87)
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.
