Understanding how galaxies form and evolve over cosmic time is one of the central questions in astrophysics. A key piece of this puzzle is galaxy morphology, since features like spiral arms, bulges, or disturbed structures are closely linked to stellar populations, environments, and kinematics However, modern sky surveys observe millions to billions of galaxies, making traditional visual classification impractical. This creates a clear need for automated and scalable methods. In this work, we use Convolutional Neural Networks, or CNNs, to build a large-scale morphological catalog using galaxy images from the DELVE survey. For training, we rely on a high-quality labeled sample from the Galaxy Zoo DECaLS project, from which we construct a robust dataset of about 98,000 galaxies classified into four main types: elliptical, lenticular, spiral, and mergers. We then compare our CNN with several well-established architectures from the literature, and find that our model achieves both higher accuracy and better computational efficiency. In particular, we reach precision levels of 97% for ellipticals, 98% for lenticulars, 99% for spirals, and 92% for mergers. Finally, we apply the trained model to previously unclassified DELVE data, producing a morphological catalog of around 30 million galaxies, covering roughly 17 square degrees down to magnitude 21.5. This demonstrates how deep learning can enable morphological studies at scales that were previously unfeasible, opening new opportunities to study galaxy evolution in the era of large surveys.