Old is gold: Restoring historical intelligence the artificial way

AI is a tool and a partner in restoring and creating artwork. Image: Shutterstock

Artificial Intelligence (AI) is fast impacting our lives, and its reach now extends far beyond our daily lives, permeating the realms of culture and heritage. It is now known to be present in conserving and restoring vintage artworks, such as paintings, music, video footage, sculptures, old X-rays, and historical forensic data. AI leverages machine learning, image processing, and deep learning techniques to aid in recovering lost masterpieces and repairing spoilt artefacts, thereby bringing in missed insights into our historical past.

A fascinating area of applied AI is the restoration and recreation of irretrievable or destroyed art of vintage quality. Researchers have been utilising machine learning to permeate black-and-white photographs of three Gustav Klimt paintings lost during World War II with vibrant use of colours. The AI deployed was powerful enough to generate realistic hues as prompted by the expert's understanding of Klimt's painting style. The result was for everyone to see. They have successfully produced a stunning replica of the original masterpiece. Amsterdam people got a chance to witness a similar marvel when the famous Rembrandt masterpiece, "The Night Watch," at the Rijksmuseum was restored using the power of AI. Rembrandt's original composition was decoded based on historical sources, allowing AI to reconstruct missing elements by analysing brushstrokes, composition and lighting. Such interweaving of skills between human expertise and AI's analytical capabilities brings new life into the artwork. AI's impact extends beyond physical restoration to preserving and enhancing fragile or damaged art. The faded manuscripts at the Fitzwilliam Museum were revitalised when the mathematicians from Cambridge University deployed image analysis algorithms for digital restoration. This significantly improved the readability and ensured the longevity of those precious documents.

The deep learning algorithms with AI can automatically restore and repair old photos by eliminating scratches, enhancing facial features, and adjusting colours. AI's creativity can also generate new art based on descriptions or existing styles. For example, Hotpot AI Art uses trained algorithms to create images from textual prompts, pushing the limits of artistic creation. However, critics argue that AI-generated art lacks authenticity and meaning due to its non-human origins. On the contrary, proponents assert that AI art is a collaborative endeavour involving human input, guidance, and interpretation.

Also read: How artificial intelligence is pushing the boundaries of art and copyright

AI has also helped recover and reconstruct damaged musical and video artworks that were once considered irrecoverable. For example, the Lost Tunes project used AI to recreate lost melodies from ancient musical notations that were unreadable to humans. The project employed a deep neural network that learned the rules and patterns of old music styles, enabling it to generate new melodies based on those notations. AI can remarkably upscale audio quality, making old recordings more evident, fuller, and lifelike. For example, Neural-Love's online platform offers a service that enhances and upscales audio using AI, which boosts audio sample rates up to kHz for high-quality sound. It can also improve voices online by reducing background noise or removing vocals from audio files.