Based on a database of images, this study intends to distinguish morphologically close wood species based on anatomical sections. The methods used by archaeobotanists to identify wood species from charcoal recovered from archaeological sites are based on a combination of several anatomical criteria. However, many species have very similar morphological characteristics that limit the identification of wood at the species level, especially when the wood pieces are not well preserved, broken or fragmentary. The aim of this traineeship is to develop a new methodological approach based on machine learning methods to distinguish morphologically close species. The students will dispose of a dataset encompassing approximately 4500 SEM 2D images (archaeological samples and natural samples) from South African wood (with metadata). This project stems from a collaboration between the project-team Calisto (Inria/Cemef) and the CEPAM research center (Cultures and Environments, Prehistory, Antiquity, Middle Age). Regular interactions of the applicants are foreseen with the two teams, located in Sophia Antipolis and in Nice, respectively.
Plus d’informations :
[Site web INRIA]