Automating container damage detection with the YOLO-NAS deep learning model.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Supply Chain Management

Background:

• Manual shipping container inspections are inefficient, prone to errors, and costly.
• Damaged containers pose risks to product quality, logistics, and safety.
• Automated solutions are needed for efficient port operations.

Purpose of the Study:

• To introduce and evaluate the YOLO-NAS model for automated container damage detection.
• To address the need for high-speed, high-accuracy inspection in complex port environments.
• To compare YOLO-NAS performance against other leading object detection models.

Main Methods:

• Implementation of the YOLO-NAS deep learning model for object detection.
• Application of the model to detect various types of container damage.
• Comparative analysis with YOLOv8, Roboflow 3.0, Fmask-RCNN, and MobileNetV2.

Main Results:

• YOLO-NAS achieved a mean average precision (mAP) of 91.2%, precision of 92.4%, and recall of 84.1%.
• YOLO-NAS demonstrated superior performance compared to YOLOv8 and Roboflow 3.0.
• YOLO-NAS offers real-time assessment capabilities crucial for port logistics, unlike other models.

Conclusions:

• YOLO-NAS is highly effective for automated container damage detection in seaports.
• The model enhances logistics efficiency, reduces costs, and improves safety.
• This technology supports the development of smart ports and predictive maintenance systems.