Estimating fruit yield is a major challenge in precision agriculture, where traditional manual counting methods are labor-intensive, error-prone, and impractical for large-scale orchards. Recent advances in computer vision and 3D imaging provide new opportunities to automate this process. This project explores a pipeline that integrates semantic segmentation with Neural Radiance Fields (NeRF) to map and analyze orchards. Segmentation models, such as Grounded-SAM2, are used to isolate fruits in complex scenes, while NeRF reconstructs detailed 3D models from multi-view images with known camera poses. Combining these techniques enables accurate visualization and provides the foundation for scalable, real-time yield estimation systems. The results highlight the potential of AI-driven tools to improve efficiency, reduce costs, and support data-driven decision-making in agriculture.