BLOG
14.11.20

In August 2020, The Verge published a fascinating article about the work of designer Daniel Voshart. Described as “a quarantine project that got a bit out of hand”, Voshart had used AI and Photoshop to bring ancient Roman emperors back to life.


Thankfully, we’re not talking Night at the Museum here (although that would be quite an accomplishment!), but the transformation of sculptures into photorealistic faces.

The article explains:


“To create his portraits, Voshart uses a combination of different software and sources. The main tool is an online program named ArtBreeder, which uses a machine learning method known as a generative adversarial network (or GAN) to manipulate portraits and landscapes…


“Voshart fed ArtBreeder images of emperors he collected from statues, coins, and paintings, and then tweaked the portraits manually based on historical descriptions, feeding them back to the GAN.”


By combining ArtBreeder with manual Photoshop edits, Voshart avoided crossing into the realm of uncanny valley – something that has plagued digital projects in the past, most notably the 2004 movie The Polar Express.


Voshart’s “artistic interpretation of an artistic interpretation” is something to behold. From Augustus and Tiberius, to Claudius and Nero, he has humanised once-cold and lifeless stone. There is warmth in their features, life in their eyes, imperfections aplenty. It’s almost as if these centuries old Emperors had sat for portraits, one sunny afternoon in July.


Voshart’s depiction of Caligula is my favourite. He looks so familiar. Like someone you work with, or a regular down the local boozer. Maybe that guy who came to service your boiler a few weeks back? It’s a testament to Voshart’s talent – and that of machine learning – that these portraits look so jaw-droppingly authentic.


From 1998 to 2004, Julian Richards hosted a BBC series entitled Meet the Ancestors. Many of the episodes culminated in facial reconstructions led by anthropologist Caroline Wilkinson (more recently responsible for the reconstructed heads of Robert the Bruce of Scotland, and the “Car Park King” himself, Richard III). Even now, the process of creating 3D reconstructions from fragments of bone is mightily impressive. From time to time, Richards would end the show stating “I almost felt as if I knew him” or words to that effect. Now, thanks to work like Voshart’s, a new generation can forge similar connections with otherwise elusive historical figures.


Game developer and artist Hu Wengu had a similar idea when utilising AI technology to bring the terracotta warriors and characters from ancient Chinese paintings to life. Self-taught on AI tools, Wengu has documented his progress via a series of YouTube videos under the alias DGSpitzer.


A short video on China Daily showcases some of his best work, which includes the restoration of old video clips using motion interpolation (generating intermediate animation frames between existing ones to make footage feel more fluid). and colourisation.


Another YouTube frequenter, Denis Shiryaev, has amassed nearly one million views for his “Top 7 world-famous portraits transformed into living human beings” video. In an 8 ½ minute video, Shiryaev tackles the Mona Lisa, Self-Portrait with Thorn Necklace and Hummingbird, The Birth of Venus, Girl with a Pearl Earring, Lady with an Ermine, American Gothic and The Night Watch.


The results are brilliant and unsettling in almost equal measure. The 500+ year-old Mona Lisa winking at the camera is something I could have happily gone my entire life without witnessing and, judging by the comments on Shiryaev’s channel, I’m far from alone:


“It's unnatural, uncanny valley level and the lack of complexion is jarring, everyone has so smooth face with no eyebags and so on. Also, what I'm seeing is people today not people back then. AI sure still has a long way to go.”


Another user added:


“These are awesome, but in all honesty: AI needs to come along a bit more before the creepy factor is gone for me. You do great work! I just think these are limited by the technology of today.”


I agree whole-heartedly. The creepy factor does not detract from the effort, talent and technological innovation that goes into such projects, but more generates an excitement for what could be possible in the future.


But it’s not just paintings that are receiving the 21st century treatment. Dinosaurs are too!

In September 2018, TechRadar published an article detailing how AI can help to restore our forgotten past. The section entitled “working out dinosaurs” is particularly illuminating:


"In uncovering the truth, sometimes AI bursts a few bubbles. Among other things it's telling us that the Tyrannosaurus Rex we know from Jurassic Park bears little resemblance to what the dinosaur was actually like.


"One current theory is that such large dinosaurs may have been feathered rather than scaly, and an AI model from the University of Manchester now suggests a T-Rex couldn’t out-run a jeep either."


It continues:


"The researchers mapped out the T-Rex’s bone and muscular structure, then used machine learning to see how fast this creature could get from point A to point B without breaking any bones.


"The findings? It’s was so big and heavy, a T-Rex could likely only walk, not run. Sprinting after some kids and scientists in search of dinner would simply put too much stress on its body."


Next they’ll be telling us that Velociraptors can’t open doors! Maybe it’s just me, but the thought of a Jurassic Park film where the T-Rex is casually strolling along with no hope in hell of catching anyone made me laugh more than I care to admit.


Thankfully not all AI advancements are creepy and / or ruin our favourite franchises.


Iconem is one such example. Founded in 2013, the start-up specialises in the 3D digitisation of endangered cultural heritage sites. According to their website:


"We work with international organisations, national governments, local authorities, and world class museums such as UNESCO, the Aga Khan Trust for Culture, the Sultanate of Oman, the City of Paris, and the Louvre. We design site-specific architectural 3D models; large-scale urban and rural 3D models; museum exhibitions; and training for local professionals."


The urgent need for such a project is encapsulated by Iconem’s mission:


"Our world’s cultural heritage is threatened. Looting, urbanisation, mass tourism, armed conflict, and climate change damage and destroy irreplaceable sites where cultures emerged, languages developed, civilisations thrived. These imperilled heritage sites embody humanity’s rich cultural diversity. It is crucial that we document and preserve them now.


"Iconem’s mission is to further conservation of these endangered places by digitising them for exploration and study, today and tomorrow. Our expert team travels the globe, combining the large-scale scanning capacity of drones and the photorealistic quality of 3D to create digital replicas of our most treasured places, record them for future generations, and champion them today."


It’s reassuring to know that such beautiful, interesting and historically significant sites can be preserved – even in digital form – for future generations to enjoy. And who knows, maybe we can enjoy a virtual visit to the Louvre to see the (less creepy) winking Mona Lisa before too long?

  • LinkedIn
  • Twitter
  • YouTube