This project, conducted in collaboration with IMAJ (Institut Mondial des Arts de la Jeunesse), focuses on the analysis of an exceptional database of more than 100,000 drawings created by young artists aged 3 to 25 from 150 countries. The collection comes from the digitization of submissions to an international annual competition organized by IMAJ in Troyes, France. The project aims to leverage contemporary algorithmic tools to explore the diversity of artistic expression within this unique corpus. In particular, it will develop a semantic analysis of the artworks using large language models (LLMs) adapted to the artistic and cultural context of the IMAJ collections. An initial phase will evaluate these models on curated subsets of the data to validate the methodology and identify limitations before scaling up the approach. The ultimate goal is to deploy this analysis at large scale to generate consistent textual interpretations of the works, thereby enriching the database with new metadata and enabling future research in digital humanities, art history, intercultural studies, and data science. Given the absence of comparable datasets, the project is inherently exploratory and addresses challenges related to the diversity of styles, techniques, cultures, and age groups represented in the collection.