A locally-developed system by the University of the Philippines Cebu (UP Cebu) integrates computer vision and deep learning techniques to grade mangoes for domestic and export markets.
The Mango Automated Neuralnet Generic Grade Assignor (MANGGA) system features a fully integrated mango grading and classification system that is capable of processing up to 800 mangoes per hour. With the increasing demand for Philippine mangoes, the system’s ability to classify mangoes efficiently and accurately streamlines the process of fruit evaluation in compliance with the Philippine National Standards.
The initial field deployment of the MANGGA system in Guba, Cebu City demonstrates its strong potential as a reliable alternative to manual sorting.
UP Cebu’s project leader, Dr. Jonnifer Sinogaya, reported that the MANGGA system can sort mangoes into three groups: small (200–249 grams [g]), medium (250–299g), and large (300–349g). The system includes three evaluation modules: 1) a size estimation model using image processing techniques that has achieved 98.05% accuracy; 2) a multi-input convolutional neural network (CNN) grading model based on general appearance that recorded a 95.4% F1-score; and 3) a CNN-based maturity detection model utilizing stem-end image that reached an accuracy of 90.62%.
By standardizing processing across batches, the system dramatically enhances postharvest mango operations, boosting efficiency, minimizing variability, and ensuring a more consistent, higher-quality product for both local and international markets. This is particularly evident in size estimation, where the MANGGA system boasts 92% accuracy compared to the manual sorter’s 45%.
The system was developed by UP Cebu Center for Environmental Informatics, in partnership with the Department of Agriculture – Region 7, Technological Institute of the Philippines, and Bureau of Plant Industry - National Mango Research and Development Center in Guimaras, through the funding support of the Philippine Council for Agriculture, Aquatic and Natural Resources Research and Development of the Department of Science and Technology (DOST-PCAARRD).
