The new exhibition at the Design Gallery shows photos from the 1920s that have been interpreted in colour by artificial intelligence until December 2023.
The past is mostly known from oral narratives, written documents and, since the end of the 19th century, from photographs and films. These imaging processes allow immediate access to events and lifeworlds. Until the 1930s, technology was only capable of producing black and white images. Certain printing processes or hand-colouring were used from the beginning to breathe colourful life into the images. However, one's own image of the past remains mostly black and white.
In an experiment in collaboration with the Fotostiftung Graubünden, the St. Moritz Documentation Library has coloured photographs from the 1920s using artificial intelligence (AI). With "Colurit", the photo foundation has developed software that makes this possible. Suddenly, Olympic champion Sonja Henje appears in a pink ice-skating costume and the postbuses in front of the railway station shine in rich yellow. The pictures from the 1920s are dedicated to themes, events and innovations that shaped the public image of St. Moritz. The exhibition shows photographs of the 1928 Olympic Games, of the railway station rebuilt in 1927 or - as an example of the famous guests - Charlie Chaplin in the snow against the backdrop of St. Moritz.
AI and the interpretation of colours by means of comparison.
Colouring black-and-white photos with AI is a technology that makes it possible to colour photos automatically without the need for human intervention. This technology is based on the so-called "deep-learning algorithms", which are able to recognise patterns and specific image motifs in black-and-white photos and colour them. This colouring is an interpretation of the historical photos based on comparisons with similar image content and statements. With the machine learning programmes, the Fotostiftung Graubünden wants to better organise its enormous flood of digital images (of more than 350,000 photos) and try to filter out the existing image content in a more targeted way. The St. Moritz Documentation Library is also trying to make its image data more accessible using AI methods. By analysing the images correctly, a programme can be trained to classify the image content correctly and assign keywords. Adding colours can highlight details and nuances in the images that were not visible in the original black and white version. This can help to improve the visual perception of past times and provide a better understanding of historical events and moments.