Glaucoma is a silent and progressive disease that can lead patients to total vision loss, so early diagnosis is essential for disease control. One of the diagnostic methods is through medical images that allow ophthalmologists to analyze the health of the optic disc. Advances in computational models can assist experts in this detection. This article aimed to employ YOLO to detect the area of interest for glaucoma detection and make annotations in the area of the optic disc and excavation. The area of interest detection model achieved a 99.5% mAP when validated on test images and an accuracy of 99% on a set of images that contained images from different databases. The detected coordinates served as the basis for cropping. The model trained to detect excavation and the optic disc in cropped images achieved a 92% mAP. This detection aided in annotating the areas and the ISNT rules annotations. Annotations based on detection were compared with annotations of the optic disc and cup areas made by experts.