Enhancing Face Recognition, Age, and Sex Estimation: From Fundamentals to Optimisation – Anders Hast

Face recognition is a widely used in everything from smartphone authentication to forensic applications, but optimising its accuracy is still an active area of research. This seminar will explore the fundamentals of face recognition, focusing on the key stages of detection, alignment, and feature vector generation. Especially, the impact of post-processing techniques on precision and recall will be examined, with a particular emphasis on historical images from the City Faces Project and some other publicly available datasets. The seminar will highlight how embedded prototype subspace classifiers enable state-of-the-art sex estimation from images. The complexities of age estimation will also be discussed, addressing the challenges that make it one of the most difficult tasks in facial analysis. Throughout the presentation, visualisation techniques, particularly t-SNE, will be used to interpret feature spaces, gain deeper insights into classification performance, and how it can be a powerful tool in data cleaning processes.

About Anders Hast