Automated and selective harvesting of fruits is increasingly vital due to high costs and seasonal labor shortages especially in advanced economies. This paper focuses on 6D pose estimation of strawberries using purely synthetic data generated through a procedural pipeline for photorealistic rendering. We employ the single-shot YOLOX-6D-Pose algorithm that leverages the YOLOX backbone (i.e., a specific deep convolution network that extracts hierarchical image features used for object detection), known for its balance between speed and accuracy, and support for edge inference. To address the lacking availability of training data, we present a flexible pipeline for generating realistic synthetic data from various 3D strawberry models via the procedural Blender pipeline, enhancing its value for training pose estimation algorithms. Quantitative evaluations show YOLOX-6D-Pose algorithm achieve comparable accuracy on both the NVIDIA RTX 3090 and Jetson Orin Nano, measured by several ADD-S metrics, which measure 6D object pose estimation accuracy by computing the average closest-point distance between model points under predicted and ground-truth poses (for symmetric objects) and evaluating it against chosen thresholds. The RTX 3090 offers superior processing speed, while the Jetson Orin Nano is ideal for resource-constrained environments, suitable for agricultural robotics. Qualitative results confirm the model’s ability to accurately infer poses of ripe and partially ripe strawberries, though challenges remain with unripe specimens. This indicates potential for future enhancements, particularly in detecting unripe strawberries by exploring color variations. The methodology can also be adapted for other fruits like apples, peaches, and plums, broadening its impact in agricultural automation.