A method for enriching a learning base intended for the training of a deep learning device. Production of an updated reference model of a dental arch and segmentation to produce a tooth model for each tooth. Acquisition of at least one updated image of the arch. For each updated image, search for virtual acquisition conditions suitable for an acquisition of a reference image of the updated reference model exhibiting a maximum match with the updated image, and acquisition of the reference image. Identification, in the reference image, of at least one reference tooth zone representing a tooth model by comparison of the updated image and the reference image determination of an updated tooth zone representing the tooth model. Assignment, to the updated tooth zone, of the tooth attribute value or values of the tooth model. Addition of the updated image enriched with value or values in the learning base.