Vineyard Vision: Investigating Mask R -CNN and YOLOv8 for Grapevine Instance Segmentation

Grapes whether as a fruit berry, wine, or raisins, are celebrated all around the world. Mostly characterized by a small round or oval shape, green to deep red color, and juicy pulp, grapes are the berries that extend their popularity from some of the well-known ancient cultures to the modern-day globalized world. This leads to grapes being grown in large quantities in vineyards with grape bunches distributed and spread across the field. Locating and monitoring these grape bunches is a very time-consuming and hectic task and requires a lot of human effort. The presented work endeavors to find an automated solution with the help of computer vision, utilizing instance segmentation and masking. The study delves into the two esteemed object detection and segmentation models, namely Mask R-CNN and YOLOv8, to identify the optimal model for the task. Comparing the two on accuracy, precision, and time, YOLOv8 with mAP 88%, takes the upper hand over Mask R-CNN with mAP 86% and establishes itself as a better reliable option for image segmentation in yards.