Object Detection

One of the most important visual perception tasks for autonomous driving is the detection of objects. Two classes of CNN-based approaches have emerged for tackling this problem: one-stage and two-stage. The two-stage approaches included an explicit step for generating region proposals, or areas of the image that are likely to contain objects. In this video, we detail how these methods work and are evaluated. The object detection experience further explores how an object detection model can be built and integrated into an autonomous system.