The project aims at developing both an engine and a methodology to update the recognition pipeline embedded into traffic monitoring cameras. Due to the broad range of real-world installation conditions, the generic recognition algorithms implemented into the cameras may be suboptimal. Therefore, a novel method for recognition performance assessment is needed. Consequently, the existing push architecture between edge devices and the central unit needs to evolve into a bidirectional architecture where the central unit will monitor the cameras and update their algorithms if needed.
Funding by INNOVIRIS
Partner : VUB: Electronics and Informatics Department (ETRO