Generic Sensor Model for Object Detection Algorithms Validation

Abstract

In the artificial intelligence era of perception algorithms being used in the state-of-the-art Advance Driver Assistance Systems, algorithm validation is not an easy task, mainly due to the amount of sensor data that must be processed at once. To provide the highest possible safety level of the solution, algorithm performance must be assessed in various difficult conditions. To address this problem, a novel Generic Sensor Model algorithm is presented, which can be directly incorporated in performance analysis of the perception algorithm in highly occluded driving scenarios. In this paper, the Generic Sensor Model algorithm is comprehensively described and its usefulness and robustness are proven with a set of experiments.