mlsanityData to 13 Oct 2026

DSLA: Dynamic smooth label assignment for efficient anchor-free object detection

arXiv 2208.00817 ↗First posted 1 Aug 2022cs.CVcited by 5

Not on the map. Why?

Abstract (arXiv)

Anchor-free detectors basically formulate object detection as dense classification and regression. For popular anchor-free detectors, it is common to introduce an individual prediction branch to estimate the quality of localization. The following inconsistencies are observed when we delve into the practices of classification and quality estimation. Firstly, for some adjacent samples which are assigned completely different labels, the trained model would produce similar classification scores. This violates the training objective and leads to performance degradation. Secondly, it is found that detected bounding boxes with higher confidences contrarily have smaller overlaps with the corresponding ground-truth. Accurately localized bounding boxes would be suppressed by less accurate ones in the Non-Maximum Suppression (NMS) procedure. To address the inconsistency problems, the Dynamic Smooth Label Assignment (DSLA) method is proposed. Based on the concept of centerness originally developed in FCOS, a smooth assignment strategy is proposed. The label is smoothed to a continuous value in [0, 1] to make a steady transition between positive and negative samples. Intersection-of-Union (IoU) is predicted dynamically during training and is coupled with the smoothed label. The dynamic smooth label is assigned to supervise the classification branch. Under such supervision, quality estimation branch is naturally merged into the classification branch, which simplifies the architecture of anchor-free detector. Comprehensive experiments are conducted on the MS COCO benchmark. It is demonstrated that, DSLA can significantly boost the detection accuracy by alleviating the above inconsistencies for anchor-free detectors. Our codes are released at https://github.com/YonghaoHe/DSLA.

Key facts. Cited by 5 papers in the map (data through 13 Oct 2026). Builds on 42 of its 55 references that are in the map. First posted on arXiv 1 Aug 2022.

Builds on

42 of 55 references.

References · 42

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,950
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,550
  3. 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,625
  4. 2017Mask R-CNNcited by 7,266
  5. 2016Feature Pyramid Networks for Object Detectioncited by 5,027
  6. 2014Fully Convolutional Networks for Semantic Segmentationcited by 4,849
  7. 2015You Only Look Once: Unified, Real-Time Object Detectioncited by 4,840
  8. 2015SSD: Single Shot MultiBox Detectorcited by 3,128
  9. 2013Rich feature hierarchies for accurate object detection and semantic segmentationcited by 2,162
  10. 2019FCOS: Fully Convolutional One-Stage Object Detectioncited by 1,575
  11. 2018Group Normalizationcited by 1,518
  12. 2017Deformable Convolutional Networkscited by 1,486
30 more
  1. 2016YOLO9000: Better, Faster, Strongercited by 1,430
  2. 2019Generalized Intersection over Union: A Metric and A Loss for Bounding Box Regressioncited by 1,362
  3. 2017Cascade R-CNN: Delving into High Quality Object Detectioncited by 1,275
  4. 2018CornerNet: Detecting Objects as Paired Keypointscited by 763
  5. 2019CenterNet: Keypoint Triplets for Object Detectioncited by 654
  6. 2017Understanding the Effective Receptive Field in Deep Convolutional Neural Networkscited by 551
  7. 2019Bridging the Gap Between Anchor-based and Anchor-free Detection via Adaptive Training Sample Selectioncited by 539
  8. 2019Distance-IoU Loss: Faster and Better Learning for Bounding Box Regressioncited by 455
  9. 2020Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detectioncited by 408
  10. 2019RepPoints: Point Set Representation for Object Detectioncited by 264
  11. 2019Libra R-CNN: Towards Balanced Learning for Object Detectioncited by 255
  12. 2021Seeing Out of tHe bOx: End-to-End Pre-training for Vision-Language Representation Learningcited by 187
  13. 2016A guide to convolution arithmetic for deep learningcited by 184
  14. 2021OTA: Optimal Transport Assignment for Object Detectioncited by 177
  15. 2018Acquisition of Localization Confidence for Accurate Object Detectioncited by 175
  16. 2020FCOS: A simple and strong anchor-free object detectorcited by 171
  17. 2019Scale-Aware Trident Networks for Object Detectioncited by 159
  18. 2019Bottom-up Object Detection by Grouping Extreme and Center Pointscited by 156
  19. 2020Probabilistic Anchor Assignment with IoU Prediction for Object Detectioncited by 150
  20. 2019Feature Selective Anchor-Free Module for Single-Shot Object Detectioncited by 139
  21. 2020Distilling Knowledge from Graph Convolutional Networkscited by 128
  22. 2020AutoAssign: Differentiable Label Assignment for Dense Object Detectioncited by 100
  23. 2018Grid R-CNNcited by 85
  24. 2019Dynamic Instance Normalization for Arbitrary Style Transfercited by 83
  25. 2020Factorizable Graph Convolutional Networkscited by 79
  26. 2019AugFPN: Improving Multi-scale Feature Learning for Object Detectioncited by 55
  27. 2019FoveaBox: Beyond Anchor-based Object Detectorcited by 37
  28. 2019IoU-aware Single-stage Object Detector for Accurate Localizationcited by 36
  29. 2019Progressive Sparse Local Attention for Video object detectioncited by 30
  30. 2022Point2Seq: Detecting 3D Objects as Sequencescited by 15

Built on by

Cited by year

Citing papers per year: 2022: 1; 2023: 1; 2024: 1; 2025: 2.

5 papers.

  1. 2025Robust Preference Optimization via Dynamic Target Marginscited by 13
  2. 2022Task-Balanced Distillation for Object Detectioncited by 5
  3. 2024Similarity Distance-Based Label Assignment for Tiny Object Detectioncited by 5
  4. 2025Few-Shot Learning in Video and 3D Object Detection: A Surveycited by 2
  5. 2023Efficient Visual Fault Detection for Freight Train Braking System via Heterogeneous Self Distillation in the Wildcited by 0