Smart Low-Light Outdoor Object Detection System based on YOLOv10 for Visual Aid

Accurate detection of objects in poorly lit environments is a crucial task, as visually impaired individuals rely on assistive technologies for safe and independent spatial navigation. This project used the YOLOv10 model to demonstrate and increase the effectiveness of a real-time object detection framework; image processing methods like noise reduction, gamma correction, and histogram equalization were implemented to improve visibility in low-light conditions. To enhance the YOLOv10 model, a set of over 7,000 low-light outdoor images was selected and annotated. The system was tested by utilizing multiple performance metrics such as precision, recall, and mean average precision, delivering satisfactory results in real-time pedestrian, vehicle, and object detection. Our model achieved an overall accuracy score of 86%, with an ultimate F1 score of 49.3%. Our results demonstrate that the YOLOv10 model is accurate and effective and can be deployed in wearable technologies such as smart glasses to further the domain of assistive technologies. Future work for this project involves the integration of advanced technologies such as haptic feedback and AR.