Open data for autonomous driving from Baidu Apollo project
Scene ParsingThe whole dataset will include RGB videos with high resolution image sequences and per pixel annotation, survey-grade dense 3D points with semantic segmentation. Our continuous collection will further add more sensors, such as stereoscopic video and panoramic images; and cover a wide range of environment, weather, and traffic conditions.Next-Gen SimulationUse of simulation environments for autonomous driving is getting a lot of attention in the automobile and computing industry. The key challenges include safety considerations that arise when an autonomous vehicle navigates the roads surrounded by other vehicles, bicycles, or pedestrians. In order to perform extensive testing and evaluations, we need to develop good simulation systems that can be used to test these vehicle not only in typical, relatively safe scenarios, but also in uncertain and dangerous environments. Our current open simulation tools include WorldSim and LogSim. We are also developing next generation technologies that can be used to generate realistic simulations of real-world traffic scenarios and driver behaviors.