ARTIFICIAL INTELLIGENCE FOR INDENTIFICATION OF WOOD AND CHARCOAL IN ARCHAELOGICAL AND PALAEOLOGICAL PERSPECTIVE

Anthracology is a robust method of studying forest stands and their transformation as a result of climatic changes or human practices, but also the uses of wood as fuel and material. This method is based on the botanical identification of wood and charcoal preserved in archaeological sites. It is based on a visual or morphometric reading, under microscopy, of the anatomical structure of the wood, preserved thanks to carbonisation. The transfer towards archaeological interpretations releaqse on the ability of the researchers to identify the taxa. While the expertise of the specialist is effective for the identification of a majority of taxa, the anatomical proximity of certain species remains a lock to the identification of other taxa with high information value. Using the latest advances and developments in AI research, and new developments, the objective of the project is to propose, via learning models, a decision support tool for the identification of taxa for which conventional methods are inoperative, even though the archaeological and paleo-environmental issues associated with their identification are numerous. By allowing the identification of key taxa, the project will provide new information on the evolution of forest stands and their uses, from the Palaeolithic to the sub-current. At the end of the project, a free labelled image database and an interoperable interface will be made available to the wood science community . It could also be used in industry for species recognition.