mlsanityData to 14 Oct 2026

AutoFlow: Learning a Better Training Set for Optical Flow

arXiv 2104.14544 ↗First posted 29 Apr 2021cs.CVcited by 102

Not on the map. Why?

Abstract (arXiv)

Synthetic datasets play a critical role in pre-training CNN models for optical flow, but they are painstaking to generate and hard to adapt to new applications. To automate the process, we present AutoFlow, a simple and effective method to render training data for optical flow that optimizes the performance of a model on a target dataset. AutoFlow takes a layered approach to render synthetic data, where the motion, shape, and appearance of each layer are controlled by learnable hyperparameters. Experimental results show that AutoFlow achieves state-of-the-art accuracy in pre-training both PWC-Net and RAFT. Our code and data are available at https://autoflow-google.github.io .

Key facts. Cited by 102 papers in the map (data through 14 Oct 2026). Builds on 34 of its 54 references that are in the map. First posted on arXiv 29 Apr 2021.

Builds on

34 of 54 references.

References · 34

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  3. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
  4. 2016Densely Connected Convolutional Networkscited by 6,720
  5. 2016The Cityscapes Dataset for Semantic Urban Scene Understandingcited by 4,444
  6. 2017CARLA: An Open Urban Driving Simulatorcited by 2,449
  7. 2020RAFT: Recurrent All-Pairs Field Transforms for Optical Flowcited by 2,415
  8. 2019RandAugment: Practical automated data augmentation with a reduced search spacecited by 2,269
  9. 2016Neural Architecture Search with Reinforcement Learningcited by 1,499
  10. 2015FlowNet: Learning Optical Flow with Convolutional Networkscited by 1,059
  11. 2017PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volumecited by 968
  12. 2015A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimationcited by 881
22 more
  1. 2016FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networkscited by 784
  2. 2020Accelerating 3D Deep Learning with PyTorch3Dcited by 784
  3. 2018AutoAugment: Learning Augmentation Policies from Datacited by 730
  4. 2016Optical Flow Estimation using a Spatial Pyramid Networkcited by 424
  5. 2016The CMA Evolution Strategy: A Tutorialcited by 377
  6. 2017Population Based Training of Neural Networkscited by 367
  7. 2016Virtual Worlds as Proxy for Multi-Object Tracking Analysiscited by 341
  8. 2017Playing for Benchmarkscited by 169
  9. 2019Iterative Residual Refinement for Joint Optical Flow and Occlusion Estimationcited by 131
  10. 2019SelFlow: Self-Supervised Learning of Optical Flowcited by 128
  11. 2019Meta-Sim: Learning to Generate Synthetic Datasetscited by 128
  12. 2020What Matters in Unsupervised Optical Flowcited by 124
  13. 2019Domain-invariant Stereo Matching Networkscited by 109
  14. 2018Models Matter, So Does Training: An Empirical Study of CNNs for Optical Flow Estimationcited by 94
  15. 2018Hierarchical Discrete Distribution Decomposition for Match Density Estimationcited by 93
  16. 2018What Makes Good Synthetic Training Data for Learning Disparity and Optical Flow Estimation?cited by 44
  17. 2020Improving 3D Object Detection through Progressive Population Based Augmentationcited by 41
  18. 2018Learning Rigidity in Dynamic Scenes with a Moving Camera for 3D Motion Field Estimationcited by 38
  19. 2019SENSE: a Shared Encoder Network for Scene-flow Estimationcited by 38
  20. 2019Learning Multi-Human Optical Flowcited by 26
  21. 2016RenderGAN: Generating Realistic Labeled Datacited by 20
  22. 2019Learning to Generate Synthetic 3D Training Data through Hybrid Gradientcited by 3

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 5; 2022: 27; 2023: 31; 2024: 16; 2025: 16; 2026: 7.

102 papers.

Sort:

By area:

  1. 2021Perceiver IO: A General Architecture for Structured Inputs & Outputscited by 629
  2. 2023CoTracker: It is Better to Track Togethercited by 522
  3. 2024Depth Pro: Sharp Monocular Metric Depth in Less Than a Secondcited by 404
  4. 2021GMFlow: Learning Optical Flow via Global Matchingcited by 327
  5. 2023Synthetic Data from Diffusion Models Improves ImageNet Classificationcited by 327
  6. 2023TAPIR: Tracking Any Point with per-frame Initialization and temporal Refinementcited by 287
  7. 2023PointOdyssey: A Large-Scale Synthetic Dataset for Long-Term Point Trackingcited by 276
  8. 2022FlowFormer: A Transformer Architecture for Optical Flowcited by 275
  9. 2022Unifying Flow, Stereo and Depth Estimationcited by 275
  10. 2022TAP-Vid: A Benchmark for Tracking Any Point in a Videocited by 263
  11. 2022Practical Stereo Matching via Cascaded Recurrent Network with Adaptive Correlationcited by 197
  12. 2023Spring: A High-Resolution High-Detail Dataset and Benchmark for Scene Flow, Optical Flow and Stereocited by 172
38 more
  1. 2022CroCo: Self-Supervised Pre-training for 3D Vision Tasks by Cross-View Completioncited by 160
  2. 2023DynamicStereo: Consistent Dynamic Depth from Stereo Videoscited by 148
  3. 2023The Surprising Effectiveness of Diffusion Models for Optical Flow and Monocular Depth Estimationcited by 116
  4. 2023FlowFormer++: Masked Cost Volume Autoencoding for Pretraining Optical Flow Estimationcited by 114
  5. 2024GaussianFlow: Splatting Gaussian Dynamics for 4D Content Creationcited by 100
  6. 2024Stereo4D: Learning How Things Move in 3D from Internet Stereo Videoscited by 89
  7. 2022CRAFT: Cross-Attentional Flow Transformer for Robust Optical Flowcited by 82
  8. 2023VideoFlow: Exploiting Temporal Cues for Multi-frame Optical Flow Estimationcited by 82
  9. 2021PDC-Net+: Enhanced Probabilistic Dense Correspondence Networkcited by 80
  10. 2022Deep Equilibrium Optical Flow Estimationcited by 64
  11. 2025AllTracker: Efficient Dense Point Tracking at High Resolutioncited by 58
  12. 2022Residual Local Feature Network for Efficient Super-Resolutioncited by 53
  13. 2022Dense Continuous-Time Optical Flow from Events and Framescited by 46
  14. 2022IFOR: Iterative Flow Minimization for Robotic Object Rearrangementcited by 42
  15. 2022Learning Pixel Trajectories with Multiscale Contrastive Random Walkscited by 36
  16. 2023Rethinking Optical Flow from Geometric Matching Consistent Perspectivecited by 33
  17. 2022Revisiting RCAN: Improved Training for Image Super-Resolutioncited by 32
  18. 2022Disentangling Architecture and Training for Optical Flowcited by 32
  19. 2022Procedural Image Programs for Representation Learningcited by 28
  20. 2022Learn2Augment: Learning to Composite Videos for Data Augmentation in Action Recognitioncited by 27
  21. 2023Learning Optical Flow from Event Camera with Rendered Datasetcited by 26
  22. 2023DriveTrack: A Benchmark for Long-Range Point Tracking in Real-World Videoscited by 25
  23. 2022Imposing Consistency for Optical Flow Estimationcited by 24
  24. 2024Motion-prior Contrast Maximization for Dense Continuous-Time Motion Estimationcited by 24
  25. 2022RealFlow: EM-based Realistic Optical Flow Dataset Generation from Videoscited by 22
  26. 2022Face Deblurring using Dual Camera Fusion on Mobile Phonescited by 18
  27. 2023Generating Images with 3D Annotations Using Diffusion Modelscited by 18
  28. 2023GAFlow: Incorporating Gaussian Attention into Optical Flowcited by 18
  29. 2023AnyFlow: Arbitrary Scale Optical Flow with Implicit Neural Representationcited by 17
  30. 2023SAMFlow: Eliminating Any Fragmentation in Optical Flow with Segment Anything Modelcited by 15
  31. 2023TinyML: Tools, Applications, Challenges, and Future Research Directionscited by 14
  32. 2023DistractFlow: Improving Optical Flow Estimation via Realistic Distractions and Pseudo-Labelingcited by 14
  33. 2023SYNAuG: Exploiting Synthetic Data for Data Imbalance Problemscited by 13
  34. 2026Track4World: Feedforward World-centric Dense 3D Tracking of All Pixelscited by 13
  35. 2023Win-Win: Training High-Resolution Vision Transformers from Two Windowscited by 12
  36. 2022AnimeRun: 2D Animation Visual Correspondence from Open Source 3D Moviescited by 11
  37. 2023Bridging the Gap: Enhancing the Utility of Synthetic Data via Post-Processing Techniquescited by 11
  38. 2024OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolationcited by 11

50 most cited of 102 All 102 (JSON)