This AI Paper Studies the Impact of Anonymization for Training Computer Vision Models with a Focus on Autonomous Vehicles Datasets