Pest identification system in rice plants

• 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,...
• Split data for model training:
  • Training: 2100 images.
  • Validation: 700 images.
  • Testing: 700 images.
• Report: click here
• DEMO: click here
• Sample data, email phanminhtai23@gmail.com to get full dataset: click here

Overviews

Predict a photo

Camera prediction

Installation

1. git clone

git clone https://github.com/phanminhtai23/Insects-Yolov8-on-Camera-with-NodeJS-ONNX.git
cd ./Insects-Yolov8-on-Camera-with-NodeJS-ONNX

2. setup

npm install

3. run

npm run dev