mlsanityData to 14 Oct 2026

D-FINE: Redefine Regression Task in DETRs as Fine-grained Distribution Refinement

arXiv 2410.13842 ↗First posted 17 Oct 2024cs.CVcited by 74

On the map: Knowledge distillation → Detection and dense prediction distillation

Abstract (arXiv)

We introduce D-FINE, a powerful real-time object detector that achieves outstanding localization precision by redefining the bounding box regression task in DETR models. D-FINE comprises two key components: Fine-grained Distribution Refinement (FDR) and Global Optimal Localization Self-Distillation (GO-LSD). FDR transforms the regression process from predicting fixed coordinates to iteratively refining probability distributions, providing a fine-grained intermediate representation that significantly enhances localization accuracy. GO-LSD is a bidirectional optimization strategy that transfers localization knowledge from refined distributions to shallower layers through self-distillation, while also simplifying the residual prediction tasks for deeper layers. Additionally, D-FINE incorporates lightweight optimizations in computationally intensive modules and operations, achieving a better balance between speed and accuracy. Specifically, D-FINE-L / X achieves 54.0% / 55.8% AP on the COCO dataset at 124 / 78 FPS on an NVIDIA T4 GPU. When pretrained on Objects365, D-FINE-L / X attains 57.1% / 59.3% AP, surpassing all existing real-time detectors. Furthermore, our method significantly enhances the performance of a wide range of DETR models by up to 5.3% AP with negligible extra parameters and training costs. Our code and pretrained models: https://github.com/Peterande/D-FINE.

Key facts. Cited by 74 papers in the map (data through 14 Oct 2026). Placed in Detection and dense prediction distillation, one of 12 directions in Knowledge distillation. Builds on 41 of its 49 references that are in the map. First posted on arXiv 17 Oct 2024.

Builds on

41 of 49 references.

Closest ancestors · 4

Same direction, newest first

  1. 2022DETRDistill: A Universal Knowledge Distillation Framework for DETR-families17 Nov
  2. 2021Distilling Object Detectors via Decoupled Features26 Mar
  3. 2021General Instance Distillation for Object Detection3 Mar
  4. 2021Localization Distillation for Dense Object Detection24 Feb

Other references · 37

  1. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  2. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  3. 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,639
  4. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
  5. 2020End-to-End Object Detection with Transformerscited by 7,414
  6. 2015You Only Look Once: Unified, Real-Time Object Detectioncited by 4,852
  7. 2020Deformable DETR: Deformable Transformers for End-to-End Object Detectioncited by 3,422
  8. 2015SSD: Single Shot MultiBox Detectorcited by 3,130
  9. 2015Fast R-CNNcited by 2,933
  10. 2014FitNets: Hints for Thin Deep Netscited by 1,653
  11. 2019FCOS: Fully Convolutional One-Stage Object Detectioncited by 1,577
  12. 2022DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detectioncited by 1,319
25 more
  1. 2016Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfercited by 1,077
  2. 2021YOLOX: Exceeding YOLO Series in 2021cited by 1,071
  3. 2022YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectorscited by 982
  4. 2019Improved Knowledge Distillation via Teacher Assistantcited by 573
  5. 2022DAB-DETR: Dynamic Anchor Boxes are Better Queries for DETRcited by 573
  6. 2023DETRs Beat YOLOs on Real-time Object Detectioncited by 533
  7. 2024YOLOv10: Real-Time End-to-End Object Detectioncited by 515
  8. 2019Be Your Own Teacher: Improve the Performance of Convolutional Neural Networks via Self Distillationcited by 497
  9. 2022DN-DETR: Accelerate DETR Training by Introducing Query DeNoisingcited by 488
  10. 2024YOLOv9: Learning What You Want to Learn Using Programmable Gradient Informationcited by 425
  11. 2021Conditional DETR for Fast Training Convergencecited by 424
  12. 2020Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detectioncited by 409
  13. 2018CrowdHuman: A Benchmark for Detecting Human in a Crowdcited by 284
  14. 2022RTMDet: An Empirical Study of Designing Real-Time Object Detectorscited by 195
  15. 2019Distilling Object Detectors with Fine-grained Feature Imitationcited by 189
  16. 2019Gaussian YOLOv3: An Accurate and Fast Object Detector Using Localization Uncertainty for Autonomous Drivingcited by 105
  17. 2020Generalized Focal Loss V2: Learning Reliable Localization Quality Estimation for Dense Object Detectioncited by 104
  18. 2024RT-DETRv2: Improved Baseline with Bag-of-Freebies for Real-Time Detection Transformercited by 91
  19. 2020Densely Guided Knowledge Distillation using Multiple Teacher Assistantscited by 83
  20. 2023Gold-YOLO: Efficient Object Detector via Gather-and-Distribute Mechanismcited by 51
  21. 2024LW-DETR: A Transformer Replacement to YOLO for Real-Time Detectioncited by 36
  22. 2022Group DETR v2: Strong Object Detector with Encoder-Decoder Pretrainingcited by 25
  23. 2021Beyond Self-Supervision: A Simple Yet Effective Network Distillation Alternative to Improve Backbonescited by 15
  24. 2022KD-DETR: Knowledge Distillation for Detection Transformer with Consistent Distillation Points Samplingcited by 13
  25. 2022A Comprehensive Study of Real-Time Object Detection Networks Across Multiple Domains: A Surveycited by 12

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 1; 2025: 22; 2026: 51.

74 papers.

Sort:

By area:

  1. 2025YOLOv12: Attention-Centric Real-Time Object Detectorscited by 305
  2. 2025RF-DETR: Neural Architecture Search for Real-Time Detection Transformerscited by 75
  3. 2024DEIM: DETR with Improved Matching for Fast Convergencecited by 68
  4. 2026Ultralytics YOLO26: Unified Real-Time End-to-End Vision Modelscited by 57
  5. 2025Real-Time Object Detection Meets DINOv3cited by 27
  6. 2025Improving Small Drone Detection Through Multi-Scale Processing and Data Augmentationcited by 7
  7. 2025Dome-DETR: DETR with Density-Oriented Feature-Query Manipulation for Efficient Tiny Object Detectioncited by 6
  8. 2025Advanced Layout Analysis Models for Doclingcited by 6
  9. 2025LINEA: Fast and Accurate Line Detection Using Scalable Transformerscited by 3
  10. 2025Edge-case Synthesis for Fisheye Object Detection: A Data-centric Perspectivecited by 3
  11. 2025Uni-Parser Technical Reportcited by 3
  12. 2026QuPAINT: Physics-Aware Instruction Tuning Approach to Quantum Material Discoverycited by 3
38 more
  1. 2026EdgeCrafter: Compact ViTs for Edge Dense Prediction via Task-Specialized Distillationcited by 3
  2. 2025MRC-DETR: An Adaptive Multi-Residual Coupled Transformer for Bare Board PCB Defect Detectioncited by 2
  3. 2025DepTR-MOT: Unveiling the Potential of Depth-Informed Trajectory Refinement for Multi-Object Trackingcited by 2
  4. 2026Real-Time Oriented Object Detection Transformer in Remote Sensing Imagescited by 2
  5. 2026A Closer Look at Cross-Domain Few-Shot Object Detection: Fine-Tuning Matters and Parallel Decoder Helpscited by 2
  6. 2026MORE: A Multilingual Document Parsing Benchmark and Evaluationcited by 2
  7. 2025ContourFormer: Real-Time Contour-Based End-to-End Instance Segmentation Transformercited by 1
  8. 2025DETRPose: Real-Time End-to-End Multi-Person Pose Estimation via Modified Transformer Decoder and Novel Denoising Keypointscited by 1
  9. 2025High-Frequency Semantics and Geometric Priors for End-to-End Detection Transformers in Challenging UAV Imagerycited by 1
  10. 2025LeAD-M3D: Leveraging Asymmetric Distillation for Real-Time Monocular 3D Detectioncited by 1
  11. 2026PARL: Position-Aware Relation Learning Network for Document Layout Analysiscited by 1
  12. 2026Le-DETR: Revisiting Real-Time Detection Transformer with Efficient Encoder Designcited by 1
  13. 2026RiO-DETR: DETR for Real-time Oriented Object Detectioncited by 1
  14. 2026FB-CLIP: Fine-Grained Zero-Shot Anomaly Detection with Foreground-Background Disentanglementcited by 1
  15. 2026WUTDet: A 100K-Scale Ship Detection Dataset and Benchmarks with Dense Small Objectscited by 1
  16. 2026Rethinking Event-Based Object Detection through Representation-Level Temporal Aggregation and Model-Level Hypergraph Reasoningcited by 1
  17. 2026RT-SDGOD: Real-Time Single-Domain Generalized Object Detectioncited by 1
  18. 2026CP-WSP: A Declarative CP-SAT Framework for Configurable Multi-Constraint Workforce Schedulingcited by 1
  19. 2026ASUMOT: Motion-Consistency-Based Asynchronous UAV Detection and Tracking with Event Camerascited by 1
  20. 2025EDocNet: Efficient Datasheet Layout Analysis Based on Focus and Global Knowledge Distillationcited by 0
  21. 2025Spatial Reasoning in Foundation Models: Benchmarking Object-Centric Spatial Understandingcited by 0
  22. 2025HierLight-YOLO: A Hierarchical and Lightweight Object Detection Network for UAV Photographycited by 0
  23. 2025FMC-DETR: Frequency-Decoupled Multi-Domain Coordination for Aerial-View Object Detectioncited by 0
  24. 2025The Urban Vision Hackathon Dataset and Models: Towards Image Annotations and Accurate Vision Models for Indian Trafficcited by 0
  25. 2025VajraV1 -- The most accurate Real Time Object Detector of the YOLO familycited by 0
  26. 2025Comparing Object Detection Models for Electrical Substation Component Mappingcited by 0
  27. 2026CoLR-Det: Collaborative Latent Restoration for Small Object Detection in Low-Resolution Remote Sensing Imagescited by 0
  28. 2026Spot-Wise Smart Parking: An Edge-Enabled Architecture with YOLOv11 and Digital Twin Integrationcited by 0
  29. 2026SARES-DEIM: Sparse Mixture-of-Experts Meets DETR for Robust SAR Ship Detectioncited by 0
  30. 2026A3-FPN: Asymptotic Content-Aware Pyramid Attention Network for Dense Visual Predictioncited by 0
  31. 2026ZoomSpec: A Physics-Guided Coarse-to-Fine Framework for Wideband Spectrum Sensingcited by 0
  32. 2026Training-Free Semantic Multi-Object Tracking with Vision-Language Modelscited by 0
  33. 2026FSDETR: Frequency-Spatial Feature Enhancement for Small Object Detectioncited by 0
  34. 2026BMD-45: A Large-Scale CCTV Vehicle Detection Dataset for Urban Traffic in Developing Citiescited by 0
  35. 2026UniPCB: A Generation-Assisted Vision-Based Measurement Framework for PCB Defect Inspectioncited by 0
  36. 2026XWOD: A Real-World Benchmark for Object Detection under Extreme Weather Conditionscited by 0
  37. 2026TinyFormer: Preserving Tiny Objects in YOLO-DETR Hybrid Real-time Detectorscited by 0
  38. 2026Bridging Multimodal Fusion and Expert Routing via Spectral Reliability Descriptors for Robust Object Detectioncited by 0

50 most cited of 74 All 74 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2024DiffuBox: Refining 3D Object Detection with Point Diffusion
  2. 2026MoEIoU: Rethinking Bounding-Box Regression as a Mixture of Experts
  3. 2026Dual-guided Hierarchical Edge Localization for Large-scale Optimal Transport Across Dimensions
  4. 2025MACE: Mixture-of-Experts Accelerated Coordinate Encoding for Large-Scale Scene Localization and Rendering
  5. 2023Video OWL-ViT: Temporally-consistent open-world localization in video
  6. 2025DiffPF: Differentiable Particle Filtering with Generative Sampling via Conditional Diffusion Models
  7. 2024DPO: Dual-Perturbation Optimization for Test-time Adaptation in 3D Object Detection
  8. 2025CLoCKDistill: Consistent Location-and-Context-aware Knowledge Distillation for DETRs