MTF as a performance indicator for AI algorithms?

The modulation-transfer function (MTF) is a fundamental optical metric to measure the quality of an imaging system. In automotive industry it used qualify camera systems for ADAS/AD. Each modern ADAS/AD system includes evaluation algorithms environment perception and decision making that are based on AI/ML methods neural networks. performance these AI measured by established metrics like Average Precision (AP) or precision-recall-curves. this article we research robustness link between metric. A series numerical experiments were performed with object detection instance segmentation (cars, pedestrians) evaluated image databases varying quality. We demonstrate strong aberrations distinct loss apparent, but subtle differences – as might arise during production tolerances does not exhibit satisfactory correlation. This calls into question how reliable current practice where produced tested end-of-line (EOL) MTF, fixed MTF thresholds camera-under-test.