Fruit Detection and Size Estimation in Orchard Environments Using Deep Learning
Application process
A completed online application must be submitted by 4.30 pm 22 September 2025. Late or incomplete applications will not be accepted. Any required supporting documentation (including references) must also be received by 4.30 pm on the closing date in order for the application to be considered.
Project number
104
Project description
This project focuses on developing a computer vision system for detecting and measuring fruit on trees using deep learning techniques. Accurate on-tree fruit detection and size estimation is vital for yield prediction and efficient harvest planning, especially in large-scale orchards where manual counting is impractical.
You will begin by adapting state-of-the-art object detection models (e.g., YOLOv5) to detect fruit in challenging conditions—such as partial occlusion, varying viewpoints, and changes in distance from the camera. The project will then explore object segmentation techniques, which aim to infer the full shape of fruits even when they are partially hidden by leaves or branches. RGB-D datasets (with depth information) will also be used to improve size estimation accuracy.
This work has real-world applications in precision agriculture and contributes to building automated, non-invasive fruit monitoring systems.
This project is for a single student.
Location
Mainly at the University.
Supervisor
Deputy Head of School, Research & International
School of Engineering and Computer Science