This presentation introduces an innovative project designed to enhance the precision of urban nuclei mapping using advanced artificial intelligence techniques. The focus is on improving the urban nuclei layer managed by the Institut Cartogràfic i Geològic de Catalunya (ICGC) through the application of convolutional neural networks (CNNs). The project involves training a CNN model on a curated collection of images featuring diverse urban nuclei. This training process enables the model to learn intricate patterns and characteristics unique to urban areas. Once the model is adequately trained, it can autonomously detect and categorize urban nuclei in new images with high accuracy.