Vision Aid: A Mobile Application Using Computer Vision and Machine Learning to Enhance Color Perception for Color Blind Users Through Multisensory Feedback

Abstract

This study proposes an artificial intelligencedriven assistive tool named VisionAid, which can achieve realtime color identification and correction, developed to assist people with color vision deficiency (CVD). The proposed system utilizes the YOLOv11 object detection model to categorize visual inputs into four color groups: red, blue, yellow, and green. A total of 2,089 labeled images were collected for training and testing the proposed system. The experimental outcome shows a maximum mean Average Precision (mAP@0.5) of 96.7% and mAP@0.5:0.95 of 89.8%, indicating excellent performance of the detection task and the convergence of the model. To make the proposed system more accessible, the trained model was incorporated into a Streamlit-based web application and further developed into a mobile application (Android APK) using AppsGeyser. The proposed system is capable of providing visual color information and optional audio assistance for better accessibility. The experimental outcome clearly validates that VisionAid is an accurate, low-latency color detection system suitable for assistive mobile applications, thereby emphasizing the effectiveness of deep learning-driven vision systems for improving accessibility and the quality of life for people with CVD.