EchoSight: Blending Deep Learning and Computer Vision for Unparalleled Navigational Support for the Visually Impaired

Our research aims to magnify mobility support for visually impaired individuals by introducing an innovative EchoSight fortified with computer vision and deep learning algorithms. Central to our design ethos is facilitating heightened independence for those often dependent on external help due to visual limitations. The EchoSight leverages the power of YOLOv3, faster CNN, and R-CNN for precise obstacle detection coupled with ultrasonic sensors. In a comparative analysis, our model demonstrated an impressive 95% accuracy rate in detecting distant obstacles, outperforming traditional aids that identify at a rate of only 70%. Furthermore, EchoSight effectively alerts users of these hazards through a multi-modal system consisting of vibration feedback, buzzer notifications, and an integrated voice assistant. Such advancements represent not only a notable progression from conventional aids but also pave the way for enhanced mobility and independence for the visually impaired community.