DUNE is a next-generation neutrino experiment designed to make precision measurements of neutrino oscillation parameters, discover potential neutrino CP violation, observe neutrinos produced in supernovas, and search for physics beyond the standard model. DUNE uses liquid argon time projection chamber (LArTPC) technology in its 40-kt far detector. LArTPC offers an excellent spatial resolution, high neutrino detection ?efficiency, and superb background rejection. Reconstruction of neutrino events in DUNE's high-resolution detectors is complicated by missing energy due to argon impurities, nonlinear detector energy response, invisible energy, complicated final states, and overlapping particles. To address these issues, neutrino events can be reconstructed directly from images of the interactions in DUNE's detectors with deep learning methods, such as Convolutional Neural Networks (CNNs). In this talk, we will focus on the development of deep-learning-based reconstruction at DUNE.