Tea leaf disease detection and identification based on YOLOv7 (YOLO-T)

Area of Science:

• Agricultural Science
• Computer Science
• Artificial Intelligence

Background:

• Manual detection of tea leaf diseases is time-consuming and impacts crop yield and quality.
• Accurate diagnosis systems are crucial for effective tea plant disease management.

Purpose of the Study:

• To develop an artificial intelligence-based solution for automated tea leaf disease detection.
• To evaluate the performance of the YOLOv7 model in identifying various tea leaf diseases.

Main Methods:

• Collected 4000 digital images of five types of tea leaf diseases from Bangladesh.
• Utilized data augmentation techniques to enhance the dataset.
• Trained and validated the YOLOv7 object detection model.

Main Results:

• YOLOv7 achieved high performance metrics: 97.3% accuracy, 96.7% precision, 96.4% recall, 98.2% mAP, and a 0.965 F1-score.
• Demonstrated superior performance compared to other detection networks like CNN, DNN, and YOLOv5.
• Successfully identified tea leaf diseases in natural scene images.

Conclusions:

• YOLOv7 offers a reliable and efficient method for detecting tea leaf diseases.
• The AI-driven approach can significantly reduce manual workload for entomologists.
• Aids in rapid disease identification, minimizing economic losses in tea cultivation.