mlsanityData to 7 Oct 2026

SMURF: Self-Teaching Multi-Frame Unsupervised RAFT with Full-Image Warping

arXiv 2105.07014 ↗First posted 14 May 2021cs.CVcited by 74

Not on the map. Why?

Abstract (arXiv)

We present SMURF, a method for unsupervised learning of optical flow that improves state of the art on all benchmarks by 36% to 40% (over the prior best method UFlow) and even outperforms several supervised approaches such as PWC-Net and FlowNet2. Our method integrates architecture improvements from supervised optical flow, i.e. the RAFT model, with new ideas for unsupervised learning that include a sequence-aware self-supervision loss, a technique for handling out-of-frame motion, and an approach for learning effectively from multi-frame video data while still only requiring two frames for inference.

Key facts. Cited by 74 papers in the map (data through 7 Oct 2026). Builds on 25 of its 42 references that are in the map. First posted on arXiv 14 May 2021.

Builds on

25 of 42 references.

References · 25

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2020RAFT: Recurrent All-Pairs Field Transforms for Optical Flowcited by 2,385
  3. 2016Instance Normalization: The Missing Ingredient for Fast Stylizationcited by 1,187
  4. 2015FlowNet: Learning Optical Flow with Convolutional Networkscited by 1,054
  5. 2017PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volumecited by 966
  6. 2015A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimationcited by 878
  7. 2016FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networkscited by 780
  8. 2016Optical Flow Estimation using a Spatial Pyramid Networkcited by 421
  9. 2018GeoNet: Unsupervised Learning of Dense Depth, Optical Flow and Camera Posecited by 341
  10. 2018Competitive Collaboration: Joint Unsupervised Learning of Depth, Camera Motion, Optical Flow and Motion Segmentationcited by 222
  11. 2017UnFlow: Unsupervised Learning of Optical Flow with a Bidirectional Census Losscited by 202
  12. 2018DF-Net: Unsupervised Joint Learning of Depth and Flow using Cross-Task Consistencycited by 148
13 more
  1. 2019SelFlow: Self-Supervised Learning of Optical Flowcited by 128
  2. 2020What Matters in Unsupervised Optical Flowcited by 124
  3. 2020Learning by Analogy: Reliable Supervision from Transformations for Unsupervised Optical Flow Estimationcited by 120
  4. 2016Back to Basics: Unsupervised Learning of Optical Flow via Brightness Constancy and Motion Smoothnesscited by 101
  5. 2017Occlusion Aware Unsupervised Learning of Optical Flowcited by 101
  6. 2017Accurate Optical Flow via Direct Cost Volume Processingcited by 66
  7. 2019DDFlow: Learning Optical Flow with Unlabeled Data Distillationcited by 63
  8. 2020Flow2Stereo: Effective Self-Supervised Learning of Optical Flow and Stereo Matchingcited by 36
  9. 2016Full Flow: Optical Flow Estimation By Global Optimization over Regular Gridscited by 34
  10. 2020Reversing the cycle: self-supervised deep stereo through enhanced monocular distillationcited by 34
  11. 2017Self-Supervised Learning for Stereo Matching with Self-Improving Abilitycited by 32
  12. 2019Unsupervised Deep Epipolar Flow for Stationary or Dynamic Scenescited by 29
  13. 2018ProFlow: Learning to Predict Optical Flowcited by 10

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 2; 2022: 17; 2023: 20; 2024: 16; 2025: 12; 2026: 7.

74 papers.

Sort:

By area:

  1. 2021Conditional Object-Centric Learning from Videocited by 247
  2. 2022Simple Unsupervised Object-Centric Learning for Complex and Naturalistic Videoscited by 139
  3. 2024BootsTAP: Bootstrapped Training for Tracking-Any-Pointcited by 115
  4. 2023FlowFormer++: Masked Cost Volume Autoencoding for Pretraining Optical Flow Estimationcited by 114
  5. 2022CRAFT: Cross-Attentional Flow Transformer for Robust Optical Flowcited by 82
  6. 2022Learning Dense and Continuous Optical Flow from an Event Cameracited by 59
  7. 2023A survey on deep learning in medical image registration: new technologies, uncertainty, evaluation metrics, and beyondcited by 55
  8. 2023MC-JEPA: A Joint-Embedding Predictive Architecture for Self-Supervised Learning of Motion and Content Featurescited by 53
  9. 2022Discovering Objects that Can Movecited by 44
  10. 2022IFOR: Iterative Flow Minimization for Robotic Object Rearrangementcited by 42
  11. 2022Learning Pixel Trajectories with Multiscale Contrastive Random Walkscited by 36
  12. 2022Disentangling Architecture and Training for Optical Flowcited by 32
38 more
  1. 2023Object Discovery from Motion-Guided Tokenscited by 25
  2. 2024Motion-prior Contrast Maximization for Dense Continuous-Time Motion Estimationcited by 24
  3. 2022RealFlow: EM-based Realistic Optical Flow Dataset Generation from Videoscited by 22
  4. 2023RM-Depth: Unsupervised Learning of Recurrent Monocular Depth in Dynamic Scenescited by 22
  5. 2022Unsupervised Multi-object Segmentation by Predicting Probable Motion Patternscited by 19
  6. 2023Unifying (Machine) Vision via Counterfactual World Modelingcited by 18
  7. 2022Boosting Object Representation Learning via Motion and Object Continuitycited by 17
  8. 2024Bring Event into RGB and LiDAR: Hierarchical Visual-Motion Fusion for Scene Flowcited by 15
  9. 2022Optical Flow Training under Limited Label Budget via Active Learningcited by 14
  10. 2022M-FUSE: Multi-frame Fusion for Scene Flow Estimationcited by 14
  11. 2023DistractFlow: Improving Optical Flow Estimation via Realistic Distractions and Pseudo-Labelingcited by 14
  12. 2024Refining Pre-Trained Motion Modelscited by 12
  13. 2024OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolationcited by 11
  14. 2022MDFlow: Unsupervised Optical Flow Learning by Reliable Mutual Knowledge Distillationcited by 10
  15. 2022Self-supervised AutoFlowcited by 10
  16. 2023GyroFlow+: Gyroscope-Guided Unsupervised Deep Homography and Optical Flow Learningcited by 10
  17. 2024Self-Supervised Any-Point Tracking by Contrastive Random Walkscited by 10
  18. 2025Scene-Centric Unsupervised Panoptic Segmentationcited by 10
  19. 2022Semi-Supervised Learning of Optical Flow by Flow Supervisorcited by 9
  20. 2023SemARFlow: Injecting Semantics into Unsupervised Optical Flow Estimation for Autonomous Drivingcited by 9
  21. 2023Unsupervised Cumulative Domain Adaptation for Foggy Scene Optical Flowcited by 9
  22. 2023Pseudo Flow Consistency for Self-Supervised 6D Object Pose Estimationcited by 9
  23. 2025Flow-Anything: Learning Real-World Optical Flow Estimation from Large-Scale Single-view Imagescited by 9
  24. 2024UnSAMFlow: Unsupervised Optical Flow Guided by Segment Anything Modelcited by 8
  25. 2025World Modeling with Probabilistic Structure Integrationcited by 7
  26. 2023Doduo: Learning Dense Visual Correspondence from Unsupervised Semantic-Aware Flowcited by 6
  27. 2024Diffeomorphic Template Registration for Atmospheric Turbulence Mitigationcited by 6
  28. 2023Unsupervised Optical Flow Estimation with Dynamic Timing Representation for Spike Cameracited by 5
  29. 2023EMR-MSF: Self-Supervised Recurrent Monocular Scene Flow Exploiting Ego-Motion Rigiditycited by 4
  30. 2024Exploring the Common Appearance-Boundary Adaptation for Nighttime Optical Flowcited by 4
  31. 2024Vector-Symbolic Architecture for Event-Based Optical Flowcited by 4
  32. 2024Adverse Weather Optical Flow: Cumulative Homogeneous-Heterogeneous Adaptationcited by 4
  33. 2025Self-Supervised Learning of Motion Concepts by Optimizing Counterfactualscited by 4
  34. 2025Taming generative video models for zero-shot optical flow extractioncited by 4
  35. 2025Motion-Refined DINOSAUR for Unsupervised Multi-Object Discoverycited by 4
  36. 2022Unsupervised Flow Refinement near Motion Boundariescited by 3
  37. 2023FlowFormer: A Transformer Architecture and Its Masked Cost Volume Autoencoding for Optical Flowcited by 3
  38. 2024Self-Supervised Multi-Frame Neural Scene Flowcited by 3

50 most cited of 74 All 74 (JSON)