自动驾驶Corner Case-视觉感知检测Corner Case
《Corner Cases for Visual Perception in Automated Driving: Some Guidance on Detection Approaches》论文笔记
- Corner Cases检测使用场景-保障行驶安全
1、In the online application,it can be used as a safety monitoring and warning system which identifies situations while they occur
车上在线的Corner Case检测,可以进行安全监控和警告系统
2、In the offline application, the corner case detector is applied to large amounts of collected data to select suitable training and relevant test data in the development of new visual perception algorithms in the laboratory
离线的Corner Case检测可以检索、生成训练数据,从而训练出更好的视觉感知算法 - Corner Cases分类和定义
由易到难可以分为5级:
- Corner Cases检索方法
针对以上各场景的corner case,当前主流方法还是针对每一个corner case,通过检索/生成/仿真 技术,生产出样本反补到业务模型,使业务模型针对corner case具有更好的鲁棒性;该方式最大问题是corner case场景无穷无尽,尤其是在object level表面的明显,在object level更像是开集检测问题;
论文中检测方法比较抽象