MIDAS - anonyMIsing DAta collection for traffic Safety
MIDAS aims to solve the problem of anonymity regarding video data collected in real traffic environments. MIDAS develops machine learning algorithms to replace sensitive information in images, so that they can be saved for future use while complying with the GDPR.
The first two research questions can be explained with the following scenario. Linnea drives her car to work every morning and her car stores video data for the developers at her research department. Every now and then, she meets her neighbour Sara who is out jogging:
RQ1: How can the generated face be maintained between frames in one video sequence?
RQ2: How can we make sure that the generated face in one video sequence is different (unique) compared with the next time Sara is captured in the video?
The third research question has to do with performance of the developed algorithms:
- RQ3: What measures can we use to guarantee that we have anonymized all personal details in a video frame?
The purpose of MIDAS is to investigate the possibility of creating anonymized but unique faces and number plates in video data to replace personal data in pictures. The results from MIDAS ensure that as much as possible of the real environment and interactions are retained in data collected in road safety-related research projects.
RISE role in project
Coordinator, WP leader