Automatic generation of ground truth for the evaluation of obstacle detection and tracking techniques
This addresses the need for efficient validation of automated vehicle safety systems, though it is incremental as it builds on existing ground-truth generation methods.
The paper tackles the problem of generating ground-truth data for evaluating obstacle detection systems in automated vehicles by proposing a novel methodology that eliminates manual annotation, requiring only sensor calibration, and provides detailed precision metrics for each data item.
As automated vehicles are getting closer to becoming a reality, it will become mandatory to be able to characterise the performance of their obstacle detection systems. This validation process requires large amounts of ground-truth data, which is currently generated by manually annotation. In this paper, we propose a novel methodology to generate ground-truth kinematics datasets for specific objects in real-world scenes. Our procedure requires no annotation whatsoever, human intervention being limited to sensors calibration. We present the recording platform which was exploited to acquire the reference data and a detailed and thorough analytical study of the propagation of errors in our procedure. This allows us to provide detailed precision metrics for each and every data item in our datasets. Finally some visualisations of the acquired data are given.