This article explores how AI can assists the geoscientist to do a faster and accurate seismic interpretation work, reducing time and costs. We have tried machine learning and deep learning approaches for addressing deep-water deposits, in the case of Complex Channels (CC). The Aggregate Channel Features (ACF) algorithm detected small segments of the CC but could not enclose the whole stratigraphy element, The latest experimental results using Regions with convolution neural networks (R-CNN) algorithm confirmed high efficiency to enclose the whole CC in a single box even more, the U-net went deeper, recognizing every detail that compound the CC, facilitating the interpretation enhancing recognition of lateral and vertical distribution of the stratigraphic element. This result allowed to include CNN's in our current AI system, available for replicability to each interpreter into Ecopetrol. Keywords: Deep-water Deposits, Machine Learning, Deep Learning, Regions with convolution neural networks, Aggregate Channel Features.