The Challenge
Sorting pears efficiently is a significant challenge in the agri-food sector. Pears are highly variable in shape, size, and color, with strict quality standards and fragile surfaces that demand precision during handling. Arinze’s Smart Pears project aims to overcome these difficulties by developing a self-learning robotic system capable of sorting and packaging pears autonomously.
To achieve this, Arinze partnered with JADS in 2022 to bring in cutting-edge data science expertise and leverage the creativity and problem-solving skills of its students. This collaboration also builds on the foundational research conducted by earlier JADS student teams, enhancing innovation in this space.
What the students did
As part of the DDFVC course, JADS students worked on a real-world application of computer vision and AI to advance the Smart Pears project. Their primary task was to design a system capable of identifying substandard pears by analyzing conveyor belt images from a sorting plant.
Using object segmentation techniques, the students developed a model to detect defects such as discoloration or deformities while ensuring the sorting process minimized damage to the fruit. Building on this foundational work, new developments include integrating neural network analysis with soft grip robotics to enhance both identification and handling processes.
JADS student Cees Maessen, under the supervision of Rogier Brussee, has continued this project, focusing on improving accuracy in distinguishing good from bad pears using AI image analysis. For his resulting thesis, he won the Unilever Research Prize!
The Result
The collaboration between JADS and Arinze has brought the Smart Pears project closer to realizing autonomous pear sorting robots. The early exploratory work by students laid the foundation for advanced systems now incorporating robotics and neural networks for precision sorting.
In 2023, René Slabbekoorn returned to JADS to share the project’s progress and inspire the next generation of students. He showcased a prototype pear sorting robot that combines soft grip technology for gentle handling with AI-driven image recognition to select quality pears accurately.
The partnership has evolved into a long-lasting cooperation, fostering innovation while providing students with practical skills and valuable insights into industrial applications of data science.