Advancing AI-Based Cocoa Seed Grading System through Video and Multimedia Data Analytics

On April 10, 2025, Dr. Supavadee Aramvith and her collaborators marked a significant milestone in the second year of a joint research and development project. Working alongside Dr. Thanyasiporn Na Nan (Aj. Tik) and a talented team from the School of Agricultural Resources and the Department of Electrical Engineering, the project focuses on developing an AI-based Cocoa Seed Quality Grading System as part of a senior engineering project.

This week, the team successfully tested Version 1 of the automatic cocoa seed grading prototype—a custom-designed cabinet integrating controlled lighting and camera setup to photograph 100 halved cocoa beans. The hardware system was jointly designed and built by Dr. Thavida Maneewarn and her team at Raina Robotech.

The imaging system connects to software powered by a deep learning model, capable of grading the beans based on color and classification criteria derived from industry standards. The system provides predictions using both internal and external quality indicators, identifying and flagging lower-grade seeds.

Additionally, the team employed the traditional cut test method for verification, and discussions are underway on whether the grading process could be optimized to eliminate the need for physical cutting. Future steps include refining the model to increase accuracy and exploring non-destructive quality assessment techniques.

This collaborative work highlights Dr. Supavadee’s commitment to applying AI and video analytics in agriculture, demonstrating the real-world impact of interdisciplinary innovation.