UP Mindanao Develops AIDurian, an AI System That Grades Durians by Sight and Sound

Researchers at the University of the Philippines Mindanao in Davao have developed an artificial intelligence system to grade durians, one of the region’s signature crops, GMA News reported on October 30. The project, called the AI-based Non-Invasive Grading System for Durian, or AIDurian, uses computer vision, acoustic analysis and deep learning to assess the fruit without cutting it open.

The stakes are significant for Mindanao growers. According to the Department of Science and Technology, the Philippines produces at least 80,000 metric tons of fresh durian a year, worth about ₱6 billion. Quality has traditionally been judged by looking at the fruit and tapping it, methods that project leader Dr. Armacheska Satina said are subjective, hard to replicate and can become a bottleneck when skilled graders are scarce during peak harvest.

AIDurian brings together two prototypes. INDAI, short for Inspection of Durian Using AI, is designed for checks in the field, while DuDONG, a desktop-based grading system, handles more detailed analysis. Together they assess ripeness, surface defects, the number of locules or segments, and shape. The team reports 80 per cent accuracy for maturity classification and acoustic ripeness, 93 per cent for defect detection and 98 per cent for locule analysis.

What it means in practice

For traders and exporters, consistent grading matters because price depends on quality, and disputes over ripeness or hidden defects can cost money at every step from farm gate to buyer. Satina said automating some grading tasks could support faster sorting, handling and marketing decisions and may cut postharvest losses. Agricultural exports were already flagged as an area with room to gain in industry briefs on AI gains for Filipino services, logistics and agri exports.

The project also shows research institutions building applied AI close to local industries rather than relying only on imported tools, in line with calls for DOST to scale homegrown Filipino AI. The 80 per cent figure for ripeness suggests the system is best used alongside experienced graders for now, and results will need to hold up across varieties, farms and seasons.

What to watch next is whether AIDurian moves from prototype to field trials with cooperatives and traders in Davao, whether DOST funds wider deployment, and whether the approach is adapted for other high-value fruit such as mangoes and pomelos.