Preserving and Restoring Heritage through AI

Preserving and Restoring Heritage through AI

Artificial Intelligence-based methods are working wonders to revive art that had been lost without a hope of restoration

The world has recently been fascinated by artworks created throughartificial intelligence (AI)-driven algorithms – especially those produced via natural language text prompts fed to online tools like DALL-E.But the media is mostly silent on howcreative AItechniques can help restore and preserve artworks which, otherwise were beyond repair.

Let’s admit it – artworks are usually fragile. Despite art being a part of human heritage, whatever have survived till date is just a miniscule fragment ofall creation produced ever since the dawn of civilisation. Artworks do get naturally ravaged by time, but they are also damaged by various external causes – war or political unrest; natural disasters like flooding, earthquake or fire; unforeseen accidents; ill-handling or bad storage and preservation; sheer human neglect – the list is endless! We must keep in mind that the impact that destruction of an artwork creates is not merely limited to its economic value; rather, it also erases a part of our cultural heritage. It is, therefore, our duty to protect significant artworks that capture the spirit of its time. And science is contributing its bit in preservation of art and restoration of any damaged artefact.

Traditional scientific methods have been employed in preserving and restoring artworks since the 18th century. The methods got more and more sophisticated as time progressed, and scientific techniques developed by leaps and bounds through the 19th and the 20th centuries. And now,artificial intelligence-based methods are working wonders to revive art that had been lost without a hope of restoration. 

Different cases, diverse usage

Let us take a quick look at the diverseadvanced techniques that are currently in use:

  • Analysis of image-basedartworks through X-rays can now be considered a traditional method. But what is new is the use of machine learning algorithms for interpreting the results of such analyses. While X-ray exposurescan revealthe sub-strata hidden beneath the layers of paint in a painting, the data derived from X-ray images are often difficult to interpret, particularly if the painting under investigation holds multiple overlapping images. This is especially the case if the artist had repainted different patterns several times. Traditional methods would be inconclusive in distinguishing such layers – but AI can effectively reconstruct the diverse phases involved in the work.
  • X-ray and CT scans are not restricted to paintings. They are also used to analyse various objects of antiquity – right from pottery fragments to statues, mummies, and fossils. For every instance, AI algorithm is enhancing the accuracy of the image analysis. Thus, AI-based toolsare also impacting archaeological heritage.
  • Another similar use is reconstruction of erased text by harnessing imaging techniques. This is allowing us to rediscover ancient textthat had been given up as lost. This comes in handy while analysing “palimpsests”. These are medieval parchment or similar writing surfaces which could be reused multiple times simply by erasing the previous text and writing over a new one. Aided with AI, modern researchers have been able to read such erased text from several layers. And that has produced invaluable results – like decoding the palimpsests of Archimedes to reveal two of his lost works!
  • Yet another marvellous achievement have been decoding the papyrus scrolls found under layers of volcanic ash at Pompeii. These were buried ever since the Vesuvius erupted to completely destroy the city in79 AD. The scrolls were intact, but so dry and brittle being singed by volcanic heat that any attempts to unroll would have crumbled them. With AI-assisted X-ray techniques, these scrolls have now been deciphered without unrolling them.
  • Similarly, a 1,700-year-old Hebrew parchment from Israel has been deciphered through text detection bya 3D-convolutional neural network.
  • AI has alsobeen successfully used in digital restoration of damaged painting, photographs, articles, and manuscripts as well as reconstructing the images on dilapidated frescoes. A process called inpaintingvia AI algorithms has been used by the MACH laboratory in Cambridge to identify damage and reconstruct lost images in old manuscripts.
  • A recent high-profile AI reconstruction was creating a projected image of the complete version ofRembrandt’s masterpiece, The Night Watch. This is significant because the painting was randomly trimmed on all four sides in 1715 when it was relocated. The AI reconstruction allowed us to surmise how the original work by the Dutch master might have looked. Likewise, a convolutional neural network was employed to reconstruct two lost panels of the famous Ghent Altarpiece by the Van Eyck brothers.
  • Onevital use of AIcan be the authentication of paintings by algorithms. Valuation of artworks is a crucial financial task,but correctly attributing works to an artist is extremely difficult. Recently, a method has been proposed to reconstruct the signature of an artist by studying the topography of the work.The process involves recording the surface height information of the work and passing it through a convolutional neural network (CNN), so that differences in brush strokes can be studied. This method also works in detecting fake artworks. In addition, algorithms to recognize the signature of a work also comes in handy.
  • Some researchers are testing the “outpainting” capability of OpenAI’s DALL-E where the tool was fedan incomplete image as input and filled in the missing portions. This was tested with damaged and deliberately erased mosaics – with very promising results.
  • Artificial intelligence can also be used to catalogue and automate tedious tasks. Researchers at Cambridge have developed an algorithm to match pottery fragments with stored profiles of potteries in a database. This helps to quickly catalogue and study the distribution of the various types of ceramics in an archaeological site.
  • DeepMind’s recently launched Ithaca, an AI model that can find missing characters in a damagedtext, is creating flutters. The model was trained on one of the largest corpora of Greek inscriptions to obtain stunning results, similar to human epigraphers.
  • Similar pattern-recognising abilities of AI has been employed to try decoding lost languages like Ugaritic or Linear B. The researchers used a model based on LSTM and embedding and obtained some interesting results.

The fantastic progress of AI in recent years has opened up interesting ways to engage increasingly powerful algorithms.However, not always do we need to develop a new algorithm for each new tasks; a more prudent – and cost-effective – approach would be to repurpose available algorithms to solvenew challenges.

Know more about the syllabus and placement record of our Top Ranked Data Science Course in KolkataData Science course in BangaloreData Science course in Hyderabad, and Data Science course in Chennai.


© 2023 Praxis. All rights reserved. | Privacy Policy
   Contact Us
Praxis Tech School
PGP in Data Science