Blind-Aid: Depth Prediction Using Object Detection to Facilitate Navigation for the Visually Impaired

Visually impaired people need help to navigate the roads and go outside without a companion. Using a walking stick does not help them much. To facilitate more effortless and a hassle-free navigation, we can use Computer Vision to detect the objects in front and find the relative distance between these objects and the user. Upon getting closer to the object, the user could be provided with a warning message to indicate the presence of the object. This helps them to become independent and walk on the roads and inside their homes easily. While existing methods use Ultrasonic sensors, the usage of Computer Vision provides more accurate results. We can detect multiple objects at the same time and provide with a warning message to the blind person in the form of an audio which contains the name of the object and also whether the object is at a safe distance from the user or if the object is too close to the user. The web application will work on YOLOv5s model. The basic structure however includes CNN layers. Once the alert message has been received in the form of text, PYTTSX3 library is used to convert the text to speech and provide the blind person with the necessary audio message. This approach is not only feasible, but also efficient and accurate, thus facilitating our vision of helping the visually impaired in the right manner.