Related Experiment Video
Updated: Jul 15, 2026
05:47
Simulation of a Scaled Assembly Process with Collaboration of a Robotic Arm and Monitoring through a Vision System for Quality Control
Published on: August 29, 2025
410
A computer vision framework for proactive anomaly detection and risk reduction in airport baggage logistics
Kalyani Vidhate1, Suraj Sawant2, Sohan Chavan3
1COEP Technological University, Pune, Maharashtra, 411005, India.
Scientific Reports
|October 31, 2025
Summary
This study introduces an AI system using YOLOv8 and GPT-4 Vision for automated airport baggage inspection, achieving high accuracy in detecting bags and handles while improving damage identification for enhanced security.
Area of Science:
- Computer Vision and Artificial Intelligence
- Airport Operations and Security
- Machine Learning for Industrial Applications
Background:
- Airport baggage handling systems face challenges in automated component detection and damage identification.
- Existing systems may lack the precision and contextual understanding required for comprehensive inspection.
- There is a need for advanced AI solutions to improve the efficiency and reliability of baggage screening.
Purpose of the Study:
- To develop and evaluate an AI-powered system for automated detection of baggage components (bags, handles, straps) and damages (cracks).
- To integrate object detection/instance segmentation (YOLOv8) with advanced reasoning (GPT-4 Vision) for comprehensive baggage inspection.
- To assess the system's performance in terms of accuracy, real-time processing, and adaptability to operational variability.
Main Methods:
- Utilized YOLOv8 for object detection and instance segmentation on a custom dataset of 2528 baggage images.
- Integrated OpenAI's GPT-4 Vision for higher-level analysis, including contextual verification and natural language damage description.
- Evaluated performance using precision, recall, F1-scores, and mean Average Precision (mAP), alongside runtime metrics (FPS, latency).
Main Results:
- Achieved high detection scores for bags (F1=0.90) and handles (F1=0.91), with good performance for damage detection (F1=0.77, mAP=0.76).
- Demonstrated robust real-time performance, with accessory detection at ~135 FPS and damage segmentation at ~62 FPS on an RTX 3090.
- The hybrid YOLOv8-GPT-4 Vision approach provided both pixel-level accuracy and semantic understanding, enhancing adaptability.
Conclusions:
- The proposed hybrid AI system sets new benchmarks for accuracy, reliability, and real-time readiness in intelligent baggage inspection.
- Highlights the importance of dataset quality, validation, and model refinement for addressing class-specific performance variations.
- The system shows significant potential for research and industrial adoption in enhancing airport security and operational efficiency.
