-, Ankit Kumar (2024) Object Detection for Visually Impaired Individuals in Different Weather Conditions. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (706kB) | Preview |
Preview |
PDF (Configuration Manual)
Download (1MB) | Preview |
Abstract
Visually impaired people use white cane to navigate through their surroundings. This research presents an object detection system designed to assist visually impaired individuals in different weather conditions. Current models do not account for real-time weather conditions. The objective of this research is to utilize the state-of-the-art YOLOv8 model and integrate weather data with it to enhance object detection accuracy and reliability under various weather conditions. The system was trained on the Pascal VOC 2012 dataset, pre-processed to simulate five major weather types: rainy, snowy, foggy, sunny, and cloudy. By incorporating real-time weather data, the model dynamically adjusts its detection parameters, providing context-aware feedback through audio using Google Text-to-Speech. The results indicate that the model performs optimally under sunny conditions, with a mean average precision (mAP@0.5) of 0.74 and an F1 score of 0.69. Performance under challenging conditions like rain and snow was lower, demonstrating the need for further optimization. The system enhances the navigational safety and independence of visually impaired users by providing real-time, intelligible audio feedback about their surroundings. This research contributes to the field of assistive technology by highlighting the importance of personalized and adaptive systems. Future work includes expanding the dataset, improving preprocessing techniques, integrating with other assistive technologies, and incorporating user feedback for continuous system improvement.
Actions (login required)
![]() |
View Item |