TinyBEV: Cross Modal Knowledge Distillation for Efficient Multi Task Bird's Eye View Perception and Planning

• URL: http://arxiv.org/abs/2509.18372v1
• Date: Mon, 22 Sep 2025 19:54:02 GMT
• Title: TinyBEV: Cross Modal Knowledge Distillation for Efficient Multi Task Bird's Eye View Perception and Planning
• Authors: Reeshad Khan, John Gauch,
• Abstract summary: We present TinyBEV, a unified, camera only Bird's Eye View (BEV) framework that distills the full-stack capabilities of a large planning-oriented teacher into a compact, real-time student model.<n>TinyBEV supports the complete autonomy stack 3D detection, HD-map segmentation, motion forecasting, occupancy prediction, and goal-directed planning within a streamlined 28M- parameter backbone.<n>Our model-agnostic, multi-stage distillation strategy combines feature-level, output-level, and adaptive region-aware supervision to effectively transfer high-capacity multi-modal knowledge to a lightweight BEV
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We present TinyBEV, a unified, camera only Bird's Eye View (BEV) framework that distills the full-stack capabilities of a large planning-oriented teacher (UniAD [19]) into a compact, real-time student model. Unlike prior efficient camera only baselines such as VAD[23] and VADv2[7], TinyBEV supports the complete autonomy stack 3D detection, HD-map segmentation, motion forecasting, occupancy prediction, and goal-directed planning within a streamlined 28M-parameter backbone, achieving a 78% reduction in parameters over UniAD [19]. Our model-agnostic, multi-stage distillation strategy combines feature-level, output-level, and adaptive region-aware supervision to effectively transfer high-capacity multi-modal knowledge to a lightweight BEV representation. On nuScenes[4], Tiny-BEV achieves 39.0 mAP for detection, 1.08 minADE for motion forecasting, and a 0.32 collision rate, while running 5x faster (11 FPS) and requiring only camera input. These results demonstrate that full-stack driving intelligence can be retained in resource-constrained settings, bridging the gap between large-scale, multi-modal perception-planning models and deployment-ready real-time autonomy.

Related papers

• InstanceBEV: Unifying Instance and BEV Representation for 3D Panoptic Segmentation [8.753189496432059]
  BEV-based 3D perception has emerged as a focal point of research in end-to-end autonomous driving.<n>We propose a novel modeling strategy, called InstanceBEV, that combines the strengths of both map-centric approaches and object-centric approaches.<n>Our method effectively extracts instance-level features within the BEV features, facilitating the implementation of global attention modeling.
  arXiv  Detail & Related papers  (2025-05-20T01:56:50Z)
• Lightweight Spatial Embedding for Vision-based 3D Occupancy Prediction [37.8001844396061]
  LightOcc is an innovative 3D occupancy prediction framework that leverages Lightweight Spatial Embedding.<n>LightOcc significantly increases the prediction accuracy of the baseline and achieves state-of-the-art performance on the Occ3D-nuScenes benchmark.
  arXiv  Detail & Related papers  (2024-12-08T15:49:35Z)
• Towards Efficient 3D Object Detection in Bird's-Eye-View Space for Autonomous Driving: A Convolutional-Only Approach [13.513005108086006]
  We propose an efficient BEV-based 3D detection framework called BEVENet. BEVENet is 3$times$ faster than contemporary state-of-the-art (SOTA) approaches on the NuScenes challenge. Our experiments show that BEVENet is 3$times$ faster than contemporary state-of-the-art (SOTA) approaches.
  arXiv  Detail & Related papers  (2023-12-01T14:52:59Z)
• QD-BEV : Quantization-aware View-guided Distillation for Multi-view 3D Object Detection [57.019527599167255]
  Multi-view 3D detection based on BEV (bird-eye-view) has recently achieved significant improvements. We show in our paper that directly applying quantization in BEV tasks will 1) make the training unstable, and 2) lead to intolerable performance degradation. Our method QD-BEV enables a novel view-guided distillation (VGD) objective, which can stabilize the quantization-aware training (QAT) while enhancing the model performance.
  arXiv  Detail & Related papers  (2023-08-21T07:06:49Z)
• BEV-IO: Enhancing Bird's-Eye-View 3D Detection with Instance Occupancy [58.926593676054424]
  We present BEV-IO, a new 3D detection paradigm to enhance BEV representation with instance occupancy information. We show that BEV-IO can outperform state-of-the-art methods while only adding a negligible increase in parameters and computational overhead.
  arXiv  Detail & Related papers  (2023-05-26T11:16:12Z)
• MetaBEV: Solving Sensor Failures for BEV Detection and Map Segmentation [104.12419434114365]
  In real-world applications, sensor corruptions and failures lead to inferior performances. We propose a robust framework, called MetaBEV, to address extreme real-world environments. We show MetaBEV outperforms prior arts by a large margin on both full and corrupted modalities.
  arXiv  Detail & Related papers  (2023-04-19T16:37:17Z)
• Fast-BEV: A Fast and Strong Bird's-Eye View Perception Baseline [76.48192454417138]
  Bird's-Eye View (BEV) representation is promising as the foundation for next-generation Autonomous Vehicle (AV) perception. This paper proposes a framework, termed Fast-BEV, which is capable of performing faster BEV perception on the on-vehicle chips.
  arXiv  Detail & Related papers  (2023-01-29T18:43:31Z)
• BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation [105.96557764248846]
  We introduce BEVFusion, a generic multi-task multi-sensor fusion framework. It unifies multi-modal features in the shared bird's-eye view representation space. It achieves 1.3% higher mAP and NDS on 3D object detection and 13.6% higher mIoU on BEV map segmentation, with 1.9x lower cost.
  arXiv  Detail & Related papers  (2022-05-26T17:59:35Z)
• M^2BEV: Multi-Camera Joint 3D Detection and Segmentation with Unified Birds-Eye View Representation [145.6041893646006]
  M$2$BEV is a unified framework that jointly performs 3D object detection and map segmentation. M$2$BEV infers both tasks with a unified model and improves efficiency.
  arXiv  Detail & Related papers  (2022-04-11T13:43:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.