A real-time rural domestic garbage detection algorithm with an improved YOLOv5s network model

Area of Science:

• Computer Science
• Environmental Science
• Artificial Intelligence

Background:

• Deep learning models are increasingly used for rural garbage classification and detection.
• Existing models suffer from high complexity, missed small targets, low accuracy, and poor real-time performance.

Purpose of the Study:

• To address limitations in current garbage detection models for rural areas.
• To develop an efficient and accurate deep learning model for rural garbage classification and detection.

Main Methods:

• An attention combination mechanism was integrated into the YOLOv5 algorithm.
• A new small object detection layer was added to the head network.
• CIoU loss function and Adam optimization algorithm were employed for model training.

Main Results:

• The proposed YOLOv5s-CSS model achieved a detection accuracy of 96.4%.
• The model processes a single garbage image in 0.021 seconds, demonstrating improved real-time performance.
• The improved algorithm shows better detection speed and accuracy compared to standard YOLOv5 and classic methods.
• Network model complexity was reduced.

Conclusions:

• The YOLOv5s-CSS model effectively overcomes the limitations of existing garbage detection systems.
• The model meets the requirements for real-time detection of domestic garbage in rural environments.
• The enhanced model offers superior performance in terms of speed, accuracy, and reduced complexity.