Raspberries, valued at $1.5 billion in 2018 face challenges in cultivation especially in monitoring growth stages. Manual monitoring is labor intensive and impractical which can lead to poor crop management and productivity issues. The use of computer vision technology, particularly, the YOLO algorithm known for its real time object detection performance has transformed agriculture by enabling efficient crop monitoring. However its application in detecting raspberries at different growth stages is still relatively unexplored. This study presents a comprehensive analysis of four YOLOv8 variations to identify and locate raspberries by their ripeness levels. It focuses on the RaspberrySet, a challenging dataset with densely overlapped raspberries from various growth stages. Through experiments involving 15 configurations of YOLOv8 we were able to determine the effective hyperparameters for this task. The results indicate that tuning specific hyperparameters significantly enhances classification accuracy with all variants of YOLOv8 outperforming the baseline model by an average of 5.91%. The YOLOv8l model emerged as the accurate among them achieving the mAP50 and mAP50 95 scores at 78.6% and 47.5%, respectively. However it struggles when it comes to identifying damaged buds due to insufficient training data. Despite its accuracy, the YOLOv8l model exhibited the longest inference time which indicate the need for faster and more efficient models suitable for deployment in agricultural environments. The next phase of research aims to develop a lightweight model that leverages YOLOv8l knowledge to enhance performance while maintaining acceptable speed.