I used GAN and a 330 BC sculpture to find out what Alexander the Great looked like

2020-08-16 · Source: Archie.AI - Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Novice, quick

Summary

A project utilized a Generative Adversarial Network (GAN) to create a lifelike portrait of Alexander the Great, combining historical sculpture with artistic interpretation. The process involved using a 330 BC bust by Leochares, specifically the "Head of Alexander" from the Acropolis Museum, as the primary facial structure input. For stylistic elements, the project incorporated the digital reconstruction by artist Jude Maris, which itself was based on the Marble head of Alexander at the Istanbul Archaeological Museum. The GAN, a machine learning framework designed by Ian Goodfellow and published in a 2014 paper, was used to generate new data with similar statistics to its training set, effectively blending the sculptural form with artistic details like heterochromia, blonde hair, and a sunburnt complexion to produce a contemporary interpretation of Alexander's appearance.

Key takeaway

For computer vision engineers or digital artists exploring historical reconstruction, consider employing Generative Adversarial Networks (GANs) to merge disparate visual sources like ancient sculptures and modern artistic interpretations. This approach allows you to synthesize a more comprehensive and lifelike representation, overcoming limitations of individual sources and potentially revealing new perspectives on historical figures.

Key insights

GANs can synthesize historical artifacts and artistic interpretations into novel visual reconstructions.

Principles

• GANs generate new data from training set statistics.
• Historical reconstructions benefit from earliest source material.

Method

Combine a 330 BC sculpture of Alexander the Great with an artist's digital reconstruction using a GAN (specifically, the open-source GAN BREEDER model) to generate a lifelike portrait.

In practice

• Use GANs for historical figure reconstruction.
• Integrate multiple visual sources for richer detail.

Topics

• Generative Adversarial Networks
• Image Synthesis
• Digital Art Reconstruction
• Historical Figure Visualization

Code references

• joel-simon/ganbreeder

Best for: Computer Vision Engineer, AI Engineer, Machine Learning Engineer, Creative Technologist

Related on AIssential

• Generative adversarial networks enable biomimetic topology fusion with balanced mechanical performance and aesthetic quality — GANs can fuse natural forms to create designs balancing mechanical performance and aesthetic appeal.
• Inceptionism with Somatic - Ep. 29 (Deep Learning SIMPLIFIED) — Inceptionism visualizes CNN learned features by applying discriminative model weights generatively to new images.
• Es Devlin’s ‘A National Portrait’ for the National Portrait Gallery — AI-powered art enables broad public participation in creating a collective, evolving national portrait.
• How to generate very large images with GANs (Ep. 76) — Multi-scale, patch-based GANs efficiently generate high-resolution images by progressively refining details across scales.
• Using design to interpret the past and envision the future — Design and AI can reconstruct lost heritage and innovate future products across diverse scales.
• Image class translation: visual inspection of class-specific hypotheticals and classification based on translation distance — Image translation networks classify by quantifying alteration needed to fit class-specific hypotheticals, offering interpretability.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Archie.AI - Medium.