Master Assignment
Reconstruction of damaged faces of Roman Emperors
Type: Master EE/CSÂ
Period: TBD
Student: (Unassigned)
If you are interested please contact :
Introduction:
Together with the University of Nijmegen, we recently started investigation of application of computer vision and face recognition techniques to analyse and recognise 2D and 3D images of roman emperors. There is a huge collection of photographs of roman emperors and other important subjects in the Roman empire available on line, in photo collections and in books. In addition, a growing collection of 3D scans of the faces (or complete statues) is becoming available. A small set of high quality 3D scans and photographs was collected by ourselves.
Figure 1: From left to right: 3D recording of Brutus, 2 photographs of roman emperors
Assignment:
Many of the statues of the emperors are damaged, some more severely than others. Sometimes only the surface has eroded, sometimes smaller or larger parts of the face are missing, e.g. the nose in the middle image above. GANs (Generative Adversarial Networks) are a type of neural networks that are used to generate natural looking images based on many examples of images of the same type. The aim of this assignment is to design and train a GAN to reconstruct the faces in photographs of damaged sculptures of roman emperors, based on a large database of photographs of roman emperors.