Camera Based Indoor Object Detection and Distance Estimation Framework for Assistive Mobility

In this paper, we propose a novel, real-time, deep learning, and computer vision-based indoor object detection and distance estimation framework for assistive mobility. Blind and visually impaired people find it difficult to deal with indoor objects and their location in their daily life. The emerging deep learning technologies can help them to do this task efficiently and conveniently by providing them with information about indoor objects present in their surroundings. In our proposed framework, we have trained the recent YOLOv7 model for indoor object detection. We have used the bounding box parameters to estimate the distance of the detected object from the user. The information on detected objects and estimated distance have been provided to visually impaired users through audio feedback. Our proposed framework will assist visually impaired people in making informed decisions and make them more confident and better prepared during their navigation in an indoor environment.