It is a system that allows identification of harmful insects on rice through images or real-time cameras. After identifying insects, the system will suggest effective preventive measures.
Written in javascript, can use client resources (use predict_vid_client.js in source) or API (use predict_vid_sever.js) for recognition.
The system uses YOLO (Yolov8n) for training, the test results on the test set are F1 ≈ 95%, mAP50 ≈ 95%.
The collected dataset includes 3,500 photos including 7 types of harmful insects in rice plants, including: Stem borer, Leaf rollers, Brown leafhoppers, Green leafhopper, Crickets, Creek compensation and Black bugs.
The images are selected and brought to the same size of 640x640. Image enhancement techniques are also used such as: image rotation, image flipping, brightness enhancement,...
