mlsanityData to 7 Oct 2026

Robust Bird's Eye View Segmentation by Adapting DINOv2

arXiv 2409.10228 ↗First posted 16 Sep 2024cs.CVcited by 6

Not on the map. Why?

Abstract (arXiv)

Extracting a Bird's Eye View (BEV) representation from multiple camera images offers a cost-effective, scalable alternative to LIDAR-based solutions in autonomous driving. However, the performance of the existing BEV methods drops significantly under various corruptions such as brightness and weather changes or camera failures. To improve the robustness of BEV perception, we propose to adapt a large vision foundational model, DINOv2, to BEV estimation using Low Rank Adaptation (LoRA). Our approach builds on the strong representation space of DINOv2 by adapting it to the BEV task in a state-of-the-art framework, SimpleBEV. Our experiments show increased robustness of BEV perception under various corruptions, with increasing gains from scaling up the model and the input resolution. We also showcase the effectiveness of the adapted representations in terms of fewer learnable parameters and faster convergence during training.

Key facts. Cited by 6 papers in the map (data through 7 Oct 2026). Builds on 29 of its 32 references that are in the map. First posted on arXiv 16 Sep 2024.

Builds on

29 of 32 references.

References · 29

  1. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
  3. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
  4. 2019nuScenes: A multimodal dataset for autonomous drivingcited by 4,485
  5. 2022BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformerscited by 1,213
  6. 2022Planning-oriented Autonomous Drivingcited by 1,068
  7. 2020Lift, Splat, Shoot: Encoding Images From Arbitrary Camera Rigs by Implicitly Unprojecting to 3Dcited by 1,040
  8. 2019Argoverse: 3D Tracking and Forecasting with Rich Mapscited by 728
  9. 2022BEVDepth: Acquisition of Reliable Depth for Multi-view 3D Object Detectioncited by 568
  10. 2023CoTracker: It is Better to Track Togethercited by 516
  11. 2019Learning by Cheatingcited by 361
  12. 2023A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondencecited by 297
17 more
  1. 2022Cross-view Transformers for real-time Map-view Semantic Segmentationcited by 264
  2. 2021End-to-End Urban Driving by Imitating a Reinforcement Learning Coachcited by 191
  3. 2024Probing the 3D Awareness of Visual Foundation Modelscited by 175
  4. 2022Learning from All Vehiclescited by 135
  5. 2022Simple-BEV: What Really Matters for Multi-Sensor BEV Perception?cited by 128
  6. 2022KING: Generating Safety-Critical Driving Scenarios for Robust Imitation via Kinematics Gradientscited by 95
  7. 2023CNOS: A Strong Baseline for CAD-based Novel Object Segmentationcited by 86
  8. 2023Self-supervised Object-Centric Learning for Videoscited by 49
  9. 2023RoboBEV: Towards Robust Bird's Eye View Perception under Corruptionscited by 42
  10. 2023MeLo: Low-rank Adaptation is Better than Fine-tuning for Medical Image Diagnosiscited by 33
  11. 2022LaRa: Latents and Rays for Multi-Camera Bird's-Eye-View Semantic Segmentationcited by 29
  12. 2024BEVCar: Camera-Radar Fusion for BEV Map and Object Segmentationcited by 27
  13. 2023PointBeV: A Sparse Approach to BeV Predictionscited by 23
  14. 2023Towards Motion Forecasting with Real-World Perception Inputs: Are End-to-End Approaches Competitive?cited by 22
  15. 2024OccFeat: Self-supervised Occupancy Feature Prediction for Pretraining BEV Segmentation Networkscited by 17
  16. 2024Can Visual Foundation Models Achieve Long-term Point Tracking?cited by 14
  17. 2024CoViS-Net: A Cooperative Visual Spatial Foundation Model for Multi-Robot Applicationscited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 5; 2026: 1.

6 papers.

  1. 2026Driving on Registerscited by 38
  2. 2025Bridging Perspectives: Foundation Model Guided BEV Maps for 3D Object Detection and Trackingcited by 9
  3. 2025Order Matters: On Parameter-Efficient Image-to-Video Probing for Recognizing Nearly Symmetric Actionscited by 3
  4. 2025FishBEV: Distortion-Resilient Bird's Eye View Segmentation with Surround-View Fisheye Camerascited by 3
  5. 2025MapFM: Foundation Model-Driven HD Mapping with Multi-Task Contextual Learningcited by 2
  6. 2025FastDINOv2: Frequency Based Curriculum Learning Improves Robustness and Training Speedcited by 2