arXiv 2508.075522025-08-12cs.CV

基于特征图质量评分的自动驾驶模型解耦功能评估

Decoupled Functional Evaluation of Autonomous Driving Models via Feature Map Quality Scoring

School of Vehicle and Mobility, Tsinghua University（车辆与移动系统学院，清华大学）；State Key Laboratory of Intelligent Green Vehicle and Mobility, Tsinghua University（智能绿色车辆与移动系统国家重点实验室，清华大学）；School of Mathematical Sciences, Nankai University（数学科学学院，南开大学）；College of Computer Science and Technology, Zhejiang University（计算机科学与技术学院，浙江大学）；SAIC-GM-Wuling Automobile Co., Ltd.（上汽通用五菱汽车有限公司）

Ludan Zhang, Sihan Wang, Yuqi Dai, Shuofei Qiao, Qinyue Luo, Lei He

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

针对端到端自动驾驶模型中间模块缺乏显式监督导致难以独立评估的问题，提出基于特征图质量评分的解耦评估方法，并构建DG-DWSS与CLIP-FMQE-Net实现实时质量分析，集成训练后在NuScenes上使NDS提升3.89%。

摘要：

端到端模型正成为自动驾驶感知与规划的主流。然而，由于缺乏对中间功能模块的显式监督信号，导致其运行机制不透明且可解释性受限，使得传统方法难以独立评估和训练这些模块。针对此问题，本研究基于特征图-真值表示相似度的评估框架，提出了一种基于特征图收敛分数（FMCS）的独立评估方法。构建了双粒度动态加权评分系统（DG-DWSS），制定了统一的量化指标——特征图质量分数，以全面评估功能模块生成的特征图质量。进一步开发了基于CLIP的特征图质量评估网络（CLIP-FMQE-Net），结合特征-真值编码器和质量分数预测头，实现对功能模块生成特征图的实时质量分析。在NuScenes数据集上的实验结果表明，将我们的评估模块集成到训练中可提升3D目标检测性能，NDS指标实现了3.89%的提升。这些结果验证了本方法在增强特征表示质量和整体模型性能方面的有效性。

英文摘要：

End-to-end models are emerging as the mainstream in autonomous driving perception and planning. However, the lack of explicit supervision signals for intermediate functional modules leads to opaque operational mechanisms and limited interpretability, making it challenging for traditional methods to independently evaluate and train these modules. Pioneering in the issue, this study builds upon the feature map-truth representation similarity-based evaluation framework and proposes an independent evaluation method based on Feature Map Convergence Score (FMCS). A Dual-Granularity Dynamic Weighted Scoring System (DG-DWSS) is constructed, formulating a unified quantitative metric - Feature Map Quality Score - to enable comprehensive evaluation of the quality of feature maps generated by functional modules. A CLIP-based Feature Map Quality Evaluation Network (CLIP-FMQE-Net) is further developed, combining feature-truth encoders and quality score prediction heads to enable real-time quality analysis of feature maps generated by functional modules. Experimental results on the NuScenes dataset demonstrate that integrating our evaluation module into the training improves 3D object detection performance, achieving a 3.89 percent gain in NDS. These results verify the effectiveness of our method in enhancing feature representation quality and overall model performance.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。