Detection and Classification of Invasive Tree Seeds in the Philippines Using YOLOv8

The spread of invasive species poses a threat to the biodiversity of native flora and fauna. Early detection of invasive species is crucial in preventing their spread and domination. This research investigates a real-time detection and classification method using deep learning and computer vision to detect and classify invasive tree seeds in the Philippines. The objective of the study is to detect and classify Philippine invasive tree species seeds in real-time using computer vision. In doing so, the specific goals must be done: (1) To create a prototype that can capture a video of a seed for detection and classification. (2) Create a dataset using seeds of the target tree species. (3) Train YOLOv8 for classifying the Philippine invasive species seeds. (4) Achieve 88% accuracy through confusion matrix evaluation. The classification in this study was assessed using a confusion matrix. The overall accuracy of the system is 97.88%. The researchers recommend expanding the dataset for invasive plants with different soil conditions and using a model with a bigger backbone on the recommended dataset.