A study of the behavior of face recognition algorithms on degraded images.

What is it about?

We study the effects of different types of image degradations on the performance of deep-learning based face recognition models to measure their robustness and provide guidelines for future research.

Featured Image

Why is it important?

Large-scale face recognition models are rarely evaluated for this kind of robustness by their original authors. We provide additional metrics for performance in the presence of image degradations and potential avenues of future research in the field of computer vision. Our findings show that the algorithms tested differ considerably in terms of robustness to different forms and intensities of image degradations, so different mitigation strategies should be considered.

Perspectives