AI-Enabled Visual Assistance: Detecting Doors and Stairs to Aid Individuals with Visual Impairments

The study aims to develop a CNN-based system to assist visually impaired individuals in navigating unfamiliar environments confidently. This system utilizes deep learning techniques to process images captured by wearable or mounted cameras. Trained on diverse datasets, the CNN identifies visual patterns for door and stair recognition, considering factors like handles, frames, and textures for doors, and step patterns, handrails, and elevation changes for stairs. Transfer learning and fine-tuning optimize the CNN architecture for robust performance in varied environments. Evaluation metrics such as precision, recall, and F1 score assess the model’s accuracy. Experimental results demonstrate the system’s effectiveness in real-time door and stair identification.