Improved Classification Approach for Fruits and Vegetables Freshness Based on Deep Learning.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Food Science

Background:

• Fruit and vegetable freshness is crucial for consumer health, purchasing decisions, and market pricing.
• Automatic classification of produce freshness using machine vision is challenging due to variations in appearance and environmental factors.

Purpose of the Study:

• To develop a deep-learning system for multiclass fruit and vegetable categorization, distinguishing between fresh and rotten items.
• To enhance the YOLOv4 model for improved accuracy and speed in fruit and vegetable freshness detection.

Main Methods:

• Developed an optimized YOLOv4 deep learning model incorporating the Mish activation function.
• Created a comprehensive image dataset of fruits and vegetables, utilizing data augmentation techniques.
• Conducted rigorous performance evaluations comparing the proposed model with existing YOLO versions.

Main Results:

• The enhanced YOLOv4 model achieved higher average precision (50.4%) compared to original YOLOv4 (49.3%) and YOLOv3 (41.7%).
• The system demonstrated precise and rapid detection capabilities for fruit and vegetable freshness classification.

Conclusions:

• The proposed deep-learning system offers a robust solution for real-time, autonomous fruit and vegetable classification in the food industry.
• This technology has significant potential to assist visually impaired individuals in selecting fresh produce and preventing foodborne illnesses.