A Novel Marker Detection System for People with Visual Impairment Using the Improved Tiny-YOLOv3 Model.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Assistive Technology

Background:

• Indoor navigation poses significant challenges for individuals with visual impairments.
• Existing solutions often fall short in providing reliable real-time assistance.

Purpose of the Study:

• To develop and evaluate a deep learning-based system for indoor navigation using markers for the visually impaired.
• To improve the accuracy and robustness of marker detection in challenging environments.

Main Methods:

• Applied deep learning, specifically an improved Tiny-YOLOv3 model, for marker detection.
• Created and augmented a dataset of marker images using various image processing techniques.
• Trained, validated, and tested the model on diverse datasets and real-world videos.

Main Results:

• The improved Tiny-YOLOv3 model achieved a 99.31% detection accuracy in challenging conditions.
• The original Tiny-YOLOv3 model achieved 96.11% accuracy.
• Modified versions of the model were implemented and compared to enhance detection capabilities.

Conclusions:

• The enhanced Tiny-YOLOv3 models demonstrate superior performance compared to the original model.
• The developed system offers a promising solution for improving indoor navigation for the visually impaired.