Accurate object detection and classification are pivotal in precision agriculture for tasks such as identifying crop varieties and assessing ripeness stages. To optimise the yield and quality of tomatoes within the variable conditions of a greenhouse, it is crucial to accurately detect and classify their ripeness levels. This study introduces YOLOv9-ResCBAM, an enhanced object detection model based on the advanced YOLOv9 architecture, designed to classify tomato ripeness levels (fully ripe, partially ripe, and unripe) in greenhouse environments. Firstly, a comprehensive tomato image dataset reflecting diverse greenhouse conditions was curated. Secondly, rigorous preprocessing techniques, including auto-orientation, resizing, and augmentation, were applied to enhance dataset quality. Finally, the proposed YOLOv9-ResCBAM model was trained and evaluated. The findings indicated that the proposed model achieved a superior mean Average Precision (mAP@0.5) of 0.912, outperforming both earlier YOLO-based detectors and other established object detection frameworks like SSD, Mask R-CNN, and Faster R-CNN. This improvement is attributed to innovations like Programmable Gradient Information (PGI), Generalised Efficient Layer Aggregation Network (GELAN), and our integration of a Residual Convolutional Block Attention Module (ResCBAM). YOLOv9-ResCBAM’s exceptional performance in accurately detecting and categorising tomatoes across ripeness stages, even in challenging greenhouse scenarios, provides a promising advancement for agricultural object detection. This study serves as a foundational step toward AI-driven solutions in crop management, enabling more efficient resource allocation and ultimately contributing to more sustainable and optimised food production practices.