将大模型的自动驾驶能力压缩到轻量级摄像头模型，实现高效端到端感知与规划。
TinyBEV: Cross Modal Knowledge Distillation for Efficient Multi Task Bird's Eye View Perception and Planning
- 通过多阶段知识蒸馏，将大模型的多模态知识迁移到轻量级BEV表示中。
- 在nuScenes上达39.0 mAP检测精度，1.08 minADE预测误差，碰撞率仅0.32。
- 仅需摄像头输入，参数减少78%，推理速度提升5倍，适合实际部署。
我们提出TinyBEV，一种统一的纯摄像头鸟瞰图（BEV）框架，将大型规划导向教师模型（UniAD[19]）的全栈能力压缩至一个紧凑的实时学生模型。与VAD[23]和VADv2[7]等先前的高效纯摄像头基线不同，TinyBEV在仅2800万参数的主干网络中支持完整自动驾驶链路：3D检测、高精地图分割、运动预测、占用预测及目标导向规划，相比UniAD[19]参数减少78%。采用模型无关的多阶段蒸馏策略，结合特征级、输出级及自适应区域感知监督，有效将高容量多模态知识迁移至轻量级BEV表示。在nuScenes[4]数据集上，TinyBEV实现39.0 mAP检测精度、1.08 minADE运动预测误差，碰撞率为0.32，运行速度达11 FPS，快5倍，且仅依赖摄像头输入。结果表明，全栈驾驶智能可在资源受限环境下得以保留，弥合了大规模多模态感知-规划模型与可部署实时自主系统之间的差距。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。