Les origines des céramiques anciennes revisitées : apport de l'apprentissage profond et des jumeaux numériques

A detailed understanding of Neolithic ceramics plays a crucial role for archaeologists, providing insights into ancient manufacturing techniques and shedding light on social connections and ancient human movements. Traditionally, the identification and classification of these ceramics heavily relied on visual methods and the empirical expertise of researchers. However, the advent of modern technologies has opened up new possibilities, such as scanning electron microscopy (SEM) and micro-tomography.
In the context of this thesis, we propose an innovative approach by harnessing Artificial Intelligence to leverage contemporary data on Neolithic ceramics. Our initial objective is to classify these ceramics based on their material composition and employed manufacturing techniques. To achieve this, we will employ deep learning, a technique that will enable us to develop a model capable of performing this classification quickly and reliably. This process will also involve a rigorous data preprocessing step to ensure the quality of results