Launched in October 2023 for three years, the HikarIA project takes its name from the association of the Japanese word for "light" (hikari 光) with the French abbreviation for artificial intelligence (IA).
The project is led by the Guimet – National Museum of Asian Arts, in partnership with TEKLIA, a French company specializing in the development of AI services for the cultural sector. HikarIA is supported by the French State, through funding from the Caisse des Dépôts, as part of the national investment plan "France 2030" and has three main goals:
Preserve, study, and share the Guimet's photographic collections;
Develop new tools for art historical research by optimizing the automated description and accessibility of historical photographs;
Enrich knowledge about photography in Japan during the Bakumatsu, Meiji and Taishō eras (1853–1926).
At the heart of the HikarIA project is an exceptional collection of photographs assembled by Dr. Joseph Dubois from the 1970s onwards and purchased by the Guimet in 2007, 2008 and 2009, thanks to the French Heritage Fund.
Comprising approximately 19 000 photographs, most of them mounted on the pages of 284 albums, the Dubois collection offers a broad overview of commercial photographic productions in Japan before the 1920s, whether in material, technical, stylistic, or even economic terms.
The collection is representative of almost all ranges of photographic products of the time: from cardboard binding to lacquer inlaid with gold and mother-of-pearl, from simple prints to hand-painted photographs, from rudimentary to the most refined colorization techniques.
The HikarIA project started with the conservation and reboxing of the entire Dubois collection, completed by the conservator Annabelle Simon in June 2025, followed by the inventory, indexing, and digitization of all the early photographs of Japan preserved at the Guimet, which represent approximately 21 000 items, by the external service provider Arkhênum.
The high-definition images and metadata produced during this phase will, on the one hand, replace the current corpus of low resolution digital images and, on the other hand, enrich the metadata collected by the historian of photography Claude Estèbe, who has researched the Dubois collection for the Guimet over the course of several missions between 2014 and 2017.
Since November 2023, TEKLIA has mobilized the Guimet's photography collections made in Japan during the Bakumatsu, Meiji and Taishō eras (1853–1926) to identify and fine-tune existing AI models able to increase the accessibility of this heritage onto this trilingual database.
The tasks entrusted to AI include:
The outlining of single photographs within digital reproductions;
The identification of all the elements visible in the photographs (places, architectures, decors, objects, people, clothing, professions, etc.);
The transcription of manuscript, typed and printed captions;
The hierarchization of descriptors along a tree structure;
The matching of similar photographs;
The geographical location of the places depicted;
The translation of the platform's content from French to Japanese and English.
Pre-existing metadata and new metadata produced by AI models will constitute a rich dataset on commercial photographic productions in Japan before 1926, able to support new detailed statistics on their iconography.
Édouard de Saint-Ours
PhD, curator of photography, Guimet Museum
Édouard holds a PhD in Art History (University of St Andrews) and Contemporary History (Université Le Havre Normandie) and is a specialist in nineteenth-century photography in Asia. His thesis, defended in 2024, examines the role of photography in the early years of French colonialism in Mainland Southeast Asia (1845-1880). He has been curator of photography at the Guimet Museum since October 2023, where he leads conservation, acquisition, research and exhibition projects on the museum's extensive photographic collections (approximately 600 000 items).
Christopher Kermorvant
PhD, CEO, TEKLIA
Christopher, President and Chief Scientist of TEKLIA, is an expert in AI for historical and cultural document analysis. A computer engineer with a PhD in AI, he conducted research at EPFL and the University of Montreal. After leading a handwriting recognition team, he founded TEKLIA in 2015 to develop AI solutions for document recognition and exploration in libraries, archives, and museums.
Mélodie Boillet
PhD, research engineer
Mélodie is specialised in Deep Learning for document processing. She holds a Master in Information Systems Architectures from INSA Rouen Normandie, a Master in Data Science and Engineering from Université de Rouen Normandie. She received a PhD in Computer Science from Université de Rouen Normandie in collaboration with the LITIS lab.
Solène Tarride
PhD, research engineer
Solène holds a PhD in Computer Science from INSA Rennes. She graduated from ENSIIE and holds a Master in Data Science and Image Processing from Université Paris-Saclay. During her PhD at IRISA, she focused on Deep Learning for historical document understanding. At TEKLIA, she is developing new methods for handwriting recognition on historical and modern documents.
Alexis Imbert
MEng, intern
Alexis holds an MEng from INSA Rouen and was an intern at TEKLIA in 2024.
Bastien Abadie
MEng, software architect
Bastien is specialised in backend development using open-source stacks and DevOps tools. He holds a MEng in Computer Science from the Ecole Nationale d'Ingénieurs de Brest. In the past he worked for various software companies: Mozilla as Senior Software engineer, Activscreen, but also as a solo consultant with his own company NextCairn.
Yoann Schneider
MEng, software engineer, ML-ops
Yoann holds a MEng from the École Nationale Supérieure d'Informatique et de Mathématiques Appliquées de Grenoble (ENSIMAG). He specializes in software development for Machine Learning and manages document processing projects.
Manon Blanco
MEng, software engineer
Manon holds an MEng from the École Nationale Supérieure d'Informatique et de Mathématiques Appliquées de Grenoble (ENSIMAG). She specializes in software development for Machine Learning projects.
Valentin Rigal
MEng, software engineer
Valentin specializes in backend development (Python), but also contributes to frontend interfaces.
Théo Lesage
Software developer
Théo is preparing a Master degree at L'Ecole by CCI. He is a software developer, with strong affinities for UX/UI work.
Saki Toriumi
PhD, associate professor, Nihon University College of Arts, Tokyo
Saki holds a PhD in Artjks from Nihon University Graduate School of Art (Tokyo). After holding various positions, including assistant curator at the Tokyo Photographic Art Museum, she is now an associate professor in the Department of Photography at Nihon University College of Art. Saki specializes in the history of photography and the study of photographic expression. She is the author of Pioneer in Modern Photography: Kanamaru Shigene (Tokyo: Kokushokankokai, 2021) and co-curator of 'Kanamaru Shigene vs. Natori Yonosuke: Olympic Photo Battle 1936' (JCII Photo Salon, 2018).
Jules Keenan
Master's Candidate, Toronto Metropolitan University
Jules Keenan holds a diploma in photography from the New Brunswick College of Craft and Design (Canada). She is currently pursuing a Master’s degree in Film and Photography Preservation and Collections Management (FPPCM) at Toronto Metropolitan University, with a focus in Photography Preservation. Previously, she served as a Photograph Archivist at the Provincial Archives of New Brunswick and is currently the archivist for the estate of the photographer Frank Grimshaw. Her research interests include the application of artificial intelligence in managing photograph collections, with a particular emphasis on automated detection of visual deterioration in photographic materials.