mlsanityData to 7 Oct 2026

OTA: Optimal Transport Assignment for Object Detection

arXiv 2103.14259 ↗First posted 26 Mar 2021cs.CVcited by 177

Not on the map. Why?

Abstract (arXiv)

Recent advances in label assignment in object detection mainly seek to independently define positive/negative training samples for each ground-truth (gt) object. In this paper, we innovatively revisit the label assignment from a global perspective and propose to formulate the assigning procedure as an Optimal Transport (OT) problem -- a well-studied topic in Optimization Theory. Concretely, we define the unit transportation cost between each demander (anchor) and supplier (gt) pair as the weighted summation of their classification and regression losses. After formulation, finding the best assignment solution is converted to solve the optimal transport plan at minimal transportation costs, which can be solved via Sinkhorn-Knopp Iteration. On COCO, a single FCOS-ResNet-50 detector equipped with Optimal Transport Assignment (OTA) can reach 40.7% mAP under 1X scheduler, outperforming all other existing assigning methods. Extensive experiments conducted on COCO and CrowdHuman further validate the effectiveness of our proposed OTA, especially its superiority in crowd scenarios. The code is available at https://github.com/Megvii-BaseDetection/OTA.

Key facts. Cited by 177 papers in the map (data through 7 Oct 2026). Builds on 46 of its 54 references that are in the map. First posted on arXiv 26 Mar 2021.

Builds on

46 of 54 references.

References · 46

  1. 2017Attention Is All You Needcited by 59,932
  2. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  3. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  4. 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,617
  5. 2020End-to-End Object Detection with Transformerscited by 7,384
  6. 2017Mask R-CNNcited by 7,259
  7. 2015You Only Look Once: Unified, Real-Time Object Detectioncited by 4,829
  8. 2015SSD: Single Shot MultiBox Detectorcited by 3,125
  9. 2015Fast R-CNNcited by 2,930
  10. 2018YOLOv3: An Incremental Improvementcited by 2,652
  11. 2016Aggregated Residual Transformations for Deep Neural Networkscited by 2,540
  12. 2013Sinkhorn Distances: Lightspeed Computation of Optimal Transportation Distancescited by 2,277
34 more
  1. 2019FCOS: Fully Convolutional One-Stage Object Detectioncited by 1,573
  2. 2016YOLO9000: Better, Faster, Strongercited by 1,429
  3. 2019Generalized Intersection over Union: A Metric and A Loss for Bounding Box Regressioncited by 1,361
  4. 2017Cascade R-CNN: Delving into High Quality Object Detectioncited by 1,275
  5. 2019Objects as Pointscited by 1,172
  6. 2018CornerNet: Detecting Objects as Paired Keypointscited by 763
  7. 2018Deformable ConvNets v2: More Deformable, Better Resultscited by 666
  8. 2019CenterNet: Keypoint Triplets for Object Detectioncited by 654
  9. 2019Bridging the Gap Between Anchor-based and Anchor-free Detection via Adaptive Training Sample Selectioncited by 539
  10. 2018CrowdHuman: A Benchmark for Detecting Human in a Crowdcited by 283
  11. 2019RepPoints: Point Set Representation for Object Detectioncited by 264
  12. 2019Libra R-CNN: Towards Balanced Learning for Object Detectioncited by 255
  13. 2016UnitBox: An Advanced Object Detection Networkcited by 205
  14. 2020Probabilistic Anchor Assignment with IoU Prediction for Object Detectioncited by 150
  15. 2017Receptive Field Block Net for Accurate and Fast Object Detectioncited by 140
  16. 2019Feature Selective Anchor-Free Module for Single-Shot Object Detectioncited by 139
  17. 2020Multi-Scale Positive Sample Refinement for Few-Shot Object Detectioncited by 116
  18. 2020AutoAssign: Differentiable Label Assignment for Dense Object Detectioncited by 100
  19. 2019Region Proposal by Guided Anchoringcited by 87
  20. 2020Cross-domain Object Detection through Coarse-to-Fine Feature Adaptationcited by 82
  21. 2020End-to-End Object Detection with Fully Convolutional Networkcited by 70
  22. 2019Adaptive NMS: Refining Pedestrian Detection in a Crowdcited by 66
  23. 2019Learning Spatial Fusion for Single-Shot Object Detectioncited by 59
  24. 2015DenseBox: Unifying Landmark Localization with End to End Object Detectioncited by 58
  25. 2019Soft Anchor-Point Object Detectioncited by 42
  26. 2020RepPoints V2: Verification Meets Regression for Object Detectioncited by 40
  27. 2020BorderDet: Border Feature for Dense Object Detectioncited by 39
  28. 2020NMS by Representative Region: Towards Crowded Pedestrian Detection by Proposal Pairingcited by 36
  29. 2019Learning from Noisy Anchors for One-stage Object Detectioncited by 35
  30. 2018MetaAnchor: Learning to Detect Objects with Customized Anchorscited by 21
  31. 2020Fine-Grained Dynamic Head for Object Detectioncited by 16
  32. 2020Rethinking Learnable Tree Filter for Generic Feature Transformcited by 10
  33. 2020PS-RCNN: Detecting Secondary Human Instances in a Crowd via Primary Object Suppressioncited by 7
  34. 2021LLA: Loss-aware Label Assignment for Dense Pedestrian Detectioncited by 7

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 14; 2022: 48; 2023: 52; 2024: 36; 2025: 19; 2026: 8.

177 papers.

Sort:

By area:

  1. 2021YOLOX: Exceeding YOLO Series in 2021cited by 1,064
  2. 2022YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectorscited by 978
  3. 2021ByteTrack: Multi-Object Tracking by Associating Every Detection Boxcited by 812
  4. 2024YOLOv9: Learning What You Want to Learn Using Programmable Gradient Informationcited by 422
  5. 2022DiffusionDet: Diffusion Model for Object Detectioncited by 354
  6. 2025YOLOv12: Attention-Centric Real-Time Object Detectorscited by 300
  7. 2022YOLOv6: A Single-Stage Object Detection Framework for Industrial Applicationscited by 273
  8. 2021TOOD: Task-aligned One-stage Object Detectioncited by 218
  9. 2022RTMDet: An Empirical Study of Designing Real-Time Object Detectorscited by 192
  10. 2023A Comprehensive Review of YOLO Architectures in Computer Vision: From YOLOv1 to YOLOv8 and YOLO-NAScited by 149
  11. 2021Oriented RepPoints for Aerial Object Detectioncited by 102
  12. 2021FCAF3D: Fully Convolutional Anchor-Free 3D Object Detectioncited by 97
38 more
  1. 2022In Defense of Online Models for Video Instance Segmentationcited by 97
  2. 2022Group DETR: Fast DETR Training with Group-Wise One-to-Many Assignmentcited by 96
  3. 2022AdaMixer: A Fast-Converging Query-Based Object Detectorcited by 78
  4. 2023Dense Distinct Query for End-to-End Object Detectioncited by 76
  5. 2023LAVA: Data Valuation without Pre-Specified Learning Algorithmscited by 69
  6. 2022Fully Convolutional One-Stage 3D Object Detection on LiDAR Range Imagescited by 56
  7. 2022Detecting tiny objects in aerial images: A normalized Wasserstein distance and a new benchmarkcited by 56
  8. 2021A Normalized Gaussian Wasserstein Distance for Tiny Object Detectioncited by 52
  9. 2022PseCo: Pseudo Labeling and Consistency Training for Semi-Supervised Object Detectioncited by 51
  10. 2022RFLA: Gaussian Receptive Field based Label Assignment for Tiny Object Detectioncited by 51
  11. 2022DiffusionInst: Diffusion Model for Instance Segmentationcited by 49
  12. 2022Equalized Focal Loss for Dense Long-Tailed Object Detectioncited by 46
  13. 2024REPARO: Compositional 3D Assets Generation with Differentiable 3D Layout Alignmentcited by 45
  14. 2022Label Matching Semi-Supervised Object Detectioncited by 44
  15. 2022Consistent-Teacher: Towards Reducing Inconsistent Pseudo-targets in Semi-supervised Object Detectioncited by 43
  16. 2023Dynamic Coarse-to-Fine Learning for Oriented Tiny Object Detectioncited by 43
  17. 2021Co-Transport for Class-Incremental Learningcited by 38
  18. 2022DAMO-YOLO : A Report on Real-Time Object Detection Designcited by 38
  19. 2022Towards Data-Efficient Detection Transformerscited by 34
  20. 2023PATS: Patch Area Transportation with Subdivision for Local Feature Matchingcited by 33
  21. 2023Transformed Distribution Matching for Missing Value Imputationcited by 31
  22. 2022UniVIP: A Unified Framework for Self-Supervised Visual Pre-trainingcited by 30
  23. 2022SoLar: Sinkhorn Label Refinery for Imbalanced Partial-Label Learningcited by 30
  24. 2022Real-time Object Detection for Streaming Perceptioncited by 29
  25. 2023Scalable Optimal Transport Methods in Machine Learning: A Contemporary Surveycited by 27
  26. 2023Semi-DETR: Semi-Supervised Object Detection with Detection Transformerscited by 27
  27. 2024AWT: Transferring Vision-Language Models via Augmentation, Weighting, and Transportationcited by 27
  28. 2022A Dual Weighting Label Assignment Scheme for Object Detectioncited by 26
  29. 2023Tuning Multi-mode Token-level Prompt Alignment across Modalitiescited by 26
  30. 2022RaLiBEV: Radar and LiDAR BEV Fusion Learning for Anchor Box Free Object Detection Systemscited by 23
  31. 2023Point2Mask: Point-supervised Panoptic Segmentation via Optimal Transportcited by 23
  32. 2021BoxeR: Box-Attention for 2D and 3D Transformerscited by 22
  33. 2023UniRef++: Segment Every Reference Object in Spatial and Temporal Spacescited by 22
  34. 2024Masked Angle-Aware Autoencoder for Remote Sensing Imagescited by 22
  35. 2022ConQueR: Query Contrast Voxel-DETR for 3D Object Detectioncited by 21
  36. 2023ByteTrackV2: 2D and 3D Multi-Object Tracking by Associating Every Detection Boxcited by 20
  37. 2021Learning to Align Sequential Actions in the Wildcited by 19
  38. 2022Localization Distillation for Object Detectioncited by 19

50 most cited of 177 All 177 (JSON)