Validation and Benchmarking for Pedestrian Video Detection based on a Sensors Simulation Platform

The evaluation stage is an important part in the validation of ADAS robustness. Moreover, the control and the repetitiveness of the experimentations were very difficult to conduct on real road due to safety reasons. Moreover, the lack of data/sensors or the complexity of the experiment are often very penalizing for a correct and exhaustive evaluation. It is for these reasons that LIVIC launched the development of a software simulation architecture (SiVIC), to support its research activities on ADAS. The use of such a simulation platform can provide both simulated sensors data and a ground truth reference for the validation stages. This paper proposes a general framework and a protocol in order to evaluate the result of pedestrian detection by camera processing.