The problem
Fresh fruit bunch grading relies on experienced visual judgement. MOPP wanted to explore computer vision as a more consistent way to assess ripeness from plantation images.
Our approach
Working with Melangking Oil Palm Plantation, we trained a three-class ripeness model using MOPP-provided images. The work established an image-based classification prototype and a foundation for future grading trials.
The engineering challenge
Recognising ripeness across changes in lighting, backgrounds, viewing angles and bunch appearance. Evaluation must distinguish genuine ripeness recognition from reliance on incidental image features.
What the work supports
AI-assisted ripeness assessment and a technical foundation for future operational grading trials, backed by a trained model and plantation data collaboration.
