mlsanityData to 14 Oct 2026

Triangular Consistency as a Universal Constraint for Learning Optical Flow

arXiv 2606.19938 ↗First posted 18 Jun 2026cs.CV cs.AIcited by 1

Not on the map. Why?

Abstract (arXiv)

We propose triangular consistency as a first-principled constraint for optical flow, which is agnostic to network architecture, supervision type, and dataset, and applies to both image-pair and multi-frame settings. This simple but powerful constraint is to compose two flows to induce a third flow and enforce consistency among the three. The composed flows may arise from (i) image pairs, yielding cycle consistency; (ii) multiple video frames, producing longer-range motion through temporal chaining; or (iii) image pairs combined with controlled synthetic transformations, which becomes data augmentation. This triangular consistency introduces negligible computational overhead and requires no additional annotations. Since it is derived directly from the geometry of optical flow, it does not rely on model-specific assumptions and serves as a ``universal'' plug-and-play component for optical flow training. Experiments show consistent improvement across supervised, unsupervised, and transfer learning settings.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 47 of its 78 references that are in the map. First posted on arXiv 18 Jun 2026.

Builds on

47 of 78 references.

References · 47

  1. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,513
  2. 2020RAFT: Recurrent All-Pairs Field Transforms for Optical Flowcited by 2,415
  3. 2015FlowNet: Learning Optical Flow with Convolutional Networkscited by 1,059
  4. 2017PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volumecited by 968
  5. 2015A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimationcited by 881
  6. 2023CoTracker: It is Better to Track Togethercited by 522
  7. 2023TAPIR: Tracking Any Point with per-frame Initialization and temporal Refinementcited by 287
  8. 2022FlowFormer: A Transformer Architecture for Optical Flowcited by 275
  9. 2019Learning Correspondence from the Cycle-Consistency of Timecited by 269
  10. 2017UnFlow: Unsupervised Learning of Optical Flow with a Bidirectional Census Losscited by 202
  11. 2023Tracking Everything Everywhere All at Oncecited by 193
  12. 2024SEA-RAFT: Simple, Efficient, Accurate RAFT for Optical Flowcited by 174
35 more
  1. 2018DF-Net: Unsupervised Joint Learning of Depth and Flow using Cross-Task Consistencycited by 148
  2. 2019SelFlow: Self-Supervised Learning of Optical Flowcited by 128
  3. 2020What Matters in Unsupervised Optical Flowcited by 124
  4. 2020Learning by Analogy: Reliable Supervision from Transformations for Unsupervised Optical Flow Estimationcited by 120
  5. 2016Back to Basics: Unsupervised Learning of Optical Flow via Brightness Constancy and Motion Smoothnesscited by 101
  6. 2017Occlusion Aware Unsupervised Learning of Optical Flowcited by 101
  7. 2019Unsupervised Depth Completion from Visual Inertial Odometrycited by 89
  8. 2022CRAFT: Cross-Attentional Flow Transformer for Robust Optical Flowcited by 82
  9. 2023VideoFlow: Exploiting Temporal Cues for Multi-frame Optical Flow Estimationcited by 82
  10. 2015EpicFlow: Edge-Preserving Interpolation of Correspondences for Optical Flowcited by 80
  11. 2021SMURF: Self-Teaching Multi-Frame Unsupervised RAFT with Full-Image Warpingcited by 74
  12. 2019DDFlow: Learning Optical Flow with Unlabeled Data Distillationcited by 63
  13. 2023MFT: Long-Term Tracking of Every Pixelcited by 62
  14. 2021Unsupervised Depth Completion with Calibrated Backprojection Layerscited by 61
  15. 2023Dense Optical Tracking: Connecting the Dotscited by 61
  16. 2022SKFlow: Learning Optical Flow with Super Kernelscited by 56
  17. 2024MemFlow: Optical Flow Estimation and Prediction with Memorycited by 55
  18. 2021Learning Optical Flow from a Few Matchescited by 53
  19. 2019Bilateral Cyclic Constraint and Adaptive Regularization for Unsupervised Monocular Depth Predictioncited by 44
  20. 2016Full Flow: Optical Flow Estimation By Global Optimization over Regular Gridscited by 34
  21. 2018Geo-Supervised Visual Depth Predictioncited by 33
  22. 2023Rethinking Optical Flow from Geometric Matching Consistent Perspectivecited by 33
  23. 2022Monitored Distillation for Positive Congruent Depth Completioncited by 24
  24. 2022Imposing Consistency for Optical Flow Estimationcited by 24
  25. 2015Flow Fields: Dense Correspondence Fields for Highly Accurate Large Displacement Optical Flow Estimationcited by 22
  26. 2020EffiScene: Efficient Per-Pixel Rigidity Inference for Unsupervised Joint Learning of Optical Flow, Depth, Camera Pose and Motion Segmentationcited by 22
  27. 2023AugUndo: Scaling Up Augmentations for Monocular Depth Completion and Estimationcited by 15
  28. 2022Optical Flow Training under Limited Label Budget via Active Learningcited by 14
  29. 2023Heteroscedastic Uncertainty Estimation Framework for Unsupervised Registrationcited by 14
  30. 2021Flow-Guided Video Inpainting with Scene Templatescited by 12
  31. 2018Conditional Prior Networks for Optical Flowcited by 11
  32. 2024LayeredFlow: A Real-World Benchmark for Non-Lambertian Multi-Layer Optical Flowcited by 11
  33. 2025FlowSeek: Optical Flow Made Easier with Depth Foundation Models and Motion Basescited by 10
  34. 2023Adaptive Correspondence Scoring for Unsupervised Medical Image Registrationcited by 9
  35. 2024Diffeomorphic Template Registration for Atmospheric Turbulence Mitigationcited by 6

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026VisTa3D: A Dataset and Benchmark for Thin Object Reconstruction from Vision, Tactile, and 3D Point Cloudscited by 1