The neural network restored the appearance of the shaman.

Serbian scientists have reconstructed the appearance and facial expressions of an ancient settler from Lepenski Vir in Serbia from his remains.

Researchers used DNA analysis, photographs of the remains and a neural network to learn what the ancient shaman's facial expressions were like.

The Lepenski Vir settlement is a Mesolithic hunting ground and one of the oldest settlements found in Europe. Archaeologists have found here about 500 remains of different people, including a skeleton that became known as a "shaman" because of the unusual lotus pose with crossed legs, in which it was found.

The shaman lived about 10,000 years ago, the researchers believe. Based on the analysis of the skeleton experts identified his height, weight and even the fact that he ate mostly seafood. To reconstruct the appearance of scientists needed to create an exact physical copy of the real skull. The researchers used photography and 3D modeling, and the finished sample of the skull was printed on a 3D printer.

In the next step, forensic scientist and archaeologist Oskar Nilsson used clay to recreate the muscle and skin layers. The researcher used information about the sex, age, ethnicity and weight of the shaman to determine the thickness of the right structures.

Once the physical reconstruction was complete, the final finish was entrusted to a neural network from Unreal Engine. The algorithm compared the appearance of the reconstructed skull with samples with similar geometry and facial proportions and reconstructed the shaman's facial expressions. It is reported that the technology uses MetaHuman Creator's extensive database of scans of real human expressions.

In the final stage, scientists clarified the appearance of their subject based on known parameters: DNA analysis showed that he had "intermediate-dark" skin, dark hair and brown eyes. The finished animated reconstruction is housed in a museum in Belgrade.