mlsanityData to 14 Oct 2026

DistillBEV: Boosting Multi-Camera 3D Object Detection with Cross-Modal Knowledge Distillation

arXiv 2309.15109 ↗First posted 26 Sep 2023cs.CV cs.ROcited by 41

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

Abstract (arXiv)

3D perception based on the representations learned from multi-camera bird's-eye-view (BEV) is trending as cameras are cost-effective for mass production in autonomous driving industry. However, there exists a distinct performance gap between multi-camera BEV and LiDAR based 3D object detection. One key reason is that LiDAR captures accurate depth and other geometry measurements, while it is notoriously challenging to infer such 3D information from merely image input. In this work, we propose to boost the representation learning of a multi-camera BEV based student detector by training it to imitate the features of a well-trained LiDAR based teacher detector. We propose effective balancing strategy to enforce the student to focus on learning the crucial features from the teacher, and generalize knowledge transfer to multi-scale layers with temporal fusion. We conduct extensive evaluations on multiple representative models of multi-camera BEV. Experiments reveal that our approach renders significant improvement over the student models, leading to the state-of-the-art performance on the popular benchmark nuScenes.

Key facts. Cited by 41 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 40 of its 46 references that are in the map. First posted on arXiv 26 Sep 2023.

Builds on

40 of 46 references.

Closest ancestors · 4

Same direction, newest first

  1. 2022BEVDistill: Cross-Modal BEV Distillation for Multi-View 3D Object Detection17 Nov
  2. 2021Focal and Global Knowledge Distillation for Detectors23 Nov
  3. 2021Instance-Conditional Knowledge Distillation for Object Detection25 Oct
  4. 2021General Instance Distillation for Object Detection3 Mar

Other references · 36

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  3. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  4. 2016PointNet: Deep Learning on Point Sets for 3D Classification and Segmentationcited by 5,248
  5. 2016Feature Pyramid Networks for Object Detectioncited by 5,036
  6. 2017Squeeze-and-Excitation Networkscited by 4,789
  7. 2019nuScenes: A multimodal dataset for autonomous drivingcited by 4,523
  8. 2018PointPillars: Fast Encoders for Object Detection from Point Cloudscited by 1,715
  9. 2014FitNets: Hints for Thin Deep Netscited by 1,653
  10. 2019FCOS: Fully Convolutional One-Stage Object Detectioncited by 1,577
  11. 2017VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detectioncited by 1,323
  12. 2022BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformerscited by 1,218
24 more
  1. 2020Center-based 3D Object Detection and Trackingcited by 1,144
  2. 2020Lift, Splat, Shoot: Encoding Images From Arbitrary Camera Rigs by Implicitly Unprojecting to 3Dcited by 1,044
  3. 2021BEVDet: High-performance Multi-camera 3D Object Detection in Bird-Eye-Viewcited by 713
  4. 2021DETR3D: 3D Object Detection from Multi-view Images via 3D-to-2D Queriescited by 658
  5. 2022BEVDepth: Acquisition of Reliable Depth for Multi-view 3D Object Detectioncited by 572
  6. 2022PETR: Position Embedding Transformation for Multi-View 3D Object Detectioncited by 552
  7. 2021FCOS3D: Fully Convolutional One-Stage Monocular 3D Object Detectioncited by 447
  8. 2019Class-balanced Grouping and Sampling for Point Cloud 3D Object Detectioncited by 350
  9. 2022BEVDet4D: Exploit Temporal Cues in Multi-camera 3D Object Detectioncited by 341
  10. 2021Is Pseudo-Lidar needed for Monocular 3D Object detection?cited by 249
  11. 2022Unifying Voxel-based Representation with Transformer for 3D Object Detectioncited by 202
  12. 2019Cross-view Semantic Segmentation for Sensing Surroundingscited by 174
  13. 2021Multimodal Virtual Point 3D Detectioncited by 166
  14. 2019An Energy and GPU-Computation Efficient Backbone Network for Real-Time Object Detectioncited by 163
  15. 2021Delving into Localization Errors for Monocular 3D Object Detectioncited by 140
  16. 2021Structured Bird's-Eye-View Traffic Scene Understanding from Onboard Imagescited by 98
  17. 2022MonoDistill: Learning Spatial Features for Monocular 3D Object Detectioncited by 68
  18. 2021Object DGCNN: 3D Object Detection using Dynamic Graphscited by 58
  19. 2021Exploring Simple 3D Multi-Object Tracking for Autonomous Drivingcited by 56
  20. 2023PillarNeXt: Rethinking Network Designs for 3D Object Detection in LiDAR Point Cloudscited by 54
  21. 2023ProphNet: Efficient Agent-Centric Motion Forecasting with Anchor-Informed Proposalscited by 48
  22. 2021Self-Supervised Pillar Motion Learning for Autonomous Drivingcited by 44
  23. 2022Hierarchical Model-Based Imitation Learning for Planning in Autonomous Drivingcited by 44
  24. 2023Transcendental Idealism of Planner: Evaluating Perception from Planning Perspective for Autonomous Drivingcited by 9

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 1; 2024: 14; 2025: 14; 2026: 12.

41 papers.

Sort:

By area:

  1. 20253DRS: MLLMs Need 3D-Aware Representation Supervision for Scene Understandingcited by 70
  2. 2024RadarDistill: Boosting Radar-based Object Detection Performance via Knowledge Distillation from LiDAR Featurescited by 28
  3. 2024CRKD: Enhanced Camera-Radar Object Detection with Cross-modality Knowledge Distillationcited by 28
  4. 2024MapDistill: Boosting Efficient Camera-based HD Map Construction via Camera-LiDAR Fusion Model Distillationcited by 26
  5. 2023Sculpting Holistic 3D Representation in Contrastive Language-Image-3D Pre-trainingcited by 18
  6. 2024OccFeat: Self-supervised Occupancy Feature Prediction for Pretraining BEV Segmentation Networkscited by 17
  7. 2024AD-H: Language-guided Autonomous Driving with Hierarchical Agentscited by 15
  8. 2024SeaBird: Segmentation in Bird's View with Dice Loss Improves Monocular 3D Detection of Large Objectscited by 12
  9. 2024MonoTAKD: Teaching Assistant Knowledge Distillation for Monocular 3D Object Detectioncited by 10
  10. 2024LabelDistill: Label-guided Cross-modal Knowledge Distillation for Camera-based 3D Object Detectioncited by 9
  11. 2024EFFOcc: Learning Efficient Occupancy Networks from Minimal Labels for Autonomous Drivingcited by 8
  12. 2025L2COcc: Lightweight Camera-Centric Semantic Scene Completion via Distillation of LiDAR Modelcited by 8
29 more
  1. 2024FSD-BEV: Foreground Self-Distillation for Multi-view 3D Object Detectioncited by 6
  2. 2025CHARM3R: Towards Unseen Camera Height Robust Monocular 3D Detectorcited by 6
  3. 2025Foundation Models for Autonomous Driving Perception: A Survey Through Core Capabilitiescited by 6
  4. 2024Cross-Modal Self-Supervised Learning with Effective Contrastive Units for LiDAR Point Cloudscited by 4
  5. 2024Unleashing the Potential of Mamba: Boosting a LiDAR 3D Sparse Detector by Using Cross-Model Knowledge Distillationcited by 4
  6. 2024TiGDistill-BEV: Multi-view BEV 3D Object Detection via Target Inner-Geometry Learning Distillationcited by 4
  7. 2026Towards Intrinsic-Aware Monocular 3D Object Detectioncited by 4
  8. 2025MapKD: Unlocking Prior Knowledge with Cross-Modal Distillation for Efficient Online HD Map Constructioncited by 3
  9. 2025RCTDistill: Cross-Modal Knowledge Distillation Framework for Radar-Camera 3D Object Detection with Temporal Fusioncited by 3
  10. 2025BridgeTA: Bridging the Representation Gap in Knowledge Distillation via Teacher Assistant for Bird's Eye View Map Segmentationcited by 2
  11. 2025KD360-VoxelBEV: LiDAR and 360-degree Camera Cross Modality Knowledge Distillation for Bird's-Eye-View Segmentationcited by 2
  12. 2025Multimodal Fusion SLAM with Fourier Attentioncited by 1
  13. 2026Learning Ego-Centric BEV Representations from a Perspective-Privileged View: Cross-View Supervision for Online HD Map Constructioncited by 1
  14. 2026AerialFusionMapNet: Online HD Map Construction with Aerial-Onboard BEV Fusioncited by 1
  15. 2024Distilling Temporal Knowledge with Masked Feature Reconstruction for 3D Object Detectioncited by 0
  16. 2025PDM-SSD: Single-Stage Three-Dimensional Object Detector With Point Dilationcited by 0
  17. 2025Expanding-and-Shrinking Binary Neural Networkscited by 0
  18. 2025Energy-Efficient Autonomous Driving with Adaptive Perception and Robust Decisioncited by 0
  19. 2025Distilling Future Temporal Knowledge with Masked Feature Reconstruction for 3D Object Detectioncited by 0
  20. 2025IMKD: Intensity-Aware Multi-Level Knowledge Distillation for Camera-Radar Fusioncited by 0
  21. 2026LiDAR-Anchored Collaborative Distillation for Robust 2D Representationscited by 0
  22. 2026Bandwidth-adaptive Cloud-Assisted 360-Degree 3D Perception for Autonomous Vehiclescited by 0
  23. 2026RayD3D: Distilling Depth Knowledge Along the Ray for Robust Multi-View 3D Object Detectioncited by 0
  24. 2026Bridging the Dimensionality Gap: A Taxonomy and Survey of 2D Vision Model Adaptation for 3D Analysiscited by 0
  25. 2026LIE: LiDAR-only HD Map Construction with Intensity Enhancement via Online Knowledge Distillationcited by 0
  26. 20263D Reconstruction and Knowledge Distillation to Improve Multi-View Image Models to Explore Spike Volume Estimation in Wheatcited by 0
  27. 2026LiAuto-GeoX: Efficient Grounded Driving Transformercited by 0
  28. 2026UMSS: Towards Unsupervised Multi-modal Semantic Segmentationcited by 0
  29. 2026Shared Semantic Codebook Distillation for Unpaired Cross-Modal Medical Classificationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2022BEV-LGKD: A Unified LiDAR-Guided Knowledge Distillation Framework for BEV 3D Object Detection
  2. 2022TiG-BEV: Multi-view BEV 3D Object Detection via Target Inner-Geometry Learning
  3. 2025BEVUDA++: Geometric-aware Unsupervised Domain Adaptation for Multi-View 3D Object Detection
  4. 2022BEVUDA: Multi-geometric Space Alignments for Domain Adaptive BEV 3D Object Detection
  5. 2023X3KD: Knowledge Distillation Across Modalities, Tasks and Stages for Multi-Camera 3D Object Detection
  6. 2023UniDistill: A Universal Cross-Modality Knowledge Distillation Framework for 3D Object Detection in Bird's-Eye View
  7. 2022Structured Knowledge Distillation Towards Efficient and Compact Multi-View 3D Detection
  8. 2022Cross-Modality Knowledge Distillation Network for Monocular 3D Object Detection