Identify medicinal plants using mobile applications
A team of authors at Yersin University in Dalat has developed a mobile application called MePDetect to help people identify common medicinal plants in Vietnam with over 90% accuracy.
Application interface and recognition results. Photo: NNC
This application uses artificial intelligence technology, especially the deep learning model CNN – EfficientB0, based on more than 800 images of 41 medicinal plants, such as Dai Bi, Diep Ha Chau, and Licorice, collected from sources, such as Google Images, Flickr, and ImageNet. When users take photos of leaves in the field, the system will quickly and accurately identify the species name, display the scientific name, reference image, uses, and used parts.
MePDetect works on Android 9.0 phones, integrating the MobileNetV2 model to increase image processing speed and ensure high accuracy in identifying medicinal plants through real images. The application not only helps people identify plants accurately, contributing to the safe and effective use of medicinal herbs, but also contributes to protecting rare plant species, avoiding unfortunate mistakes.
The team is currently expanding the training dataset, improving image processing capabilities in less-than-ideal environmental conditions such as low light or low-quality images. In addition, they are developing an automatic model update system to improve accuracy when new data or new medicinal plants are discovered. The research has been published in the Can Tho University Journal of Science, No. 2A/2025.