Weeds are undesirable and encroaching plants that proliferate quickly and vie for essential resources, including space, water, nutrients, and light, which can impact both the quality and yield of crops. One of the alternatives to reduce the influence of weeds is precision weeding, which uses image sensors and computer algorithms to identify plants and classify weeds through digital images. These images are analyzed through Deep Learning algorithms, specifically with Convolutional Neural Networks (CNN), so this study analyzes the CNN YOLOv5 architecture with all its models (n, s, m, l, and x), in the corn crop (Zea mays L.) with four types of weeds, obtaining that the classification power by confusion matrix was between 96 and 97% for the maize class, the general mAP@0.5 of the models was between 0.744 and 0.758, the differences between models were less than 2%, the F1-score values were between 0.72 and 0.74 and the confidence thresholds between 0.312 and 0.433, indicating that the model that best fits our application is the YOLOv5l and YOLOv5x models.