Assessing face image quality with LSTMs
Tommy Thorsen, Kiran Bylappa Raja, Raghavendra Ramachandra, Pankaj Wasnik, Christoph Busch · 2018
Biometric authentication using ngerprints or face recognition is making its way intothe mainstream, and there is an urgent need to make these authentication methods assecure and reliable as possible. One way to achieve better performance with a biometricauthentication method, is to introduce a quality estimation step early in the pipeline, sothat unsuitable, or low-quality samples can be rejected.While existing work predominantly focuses on algorithms for detecting specic propertiesof the face images, we investigate whether machine learning techniques can provide ageneral way to estimate overall face image quality.We train a selection of neural network types, and discover that a type of RecurrentNeural Network (RNN) called Long Short-Term Memory (LSTM) can reliably estimateface image quality, with better performance than the bespoke algorithms.