Our project is a device designed to help People with impaired vision easily navigate their environment and avoid obstacle related injuries. It uses a camera to detect the direction of various objects in the environment as well as their proximity to the user. It has a surround sound system to convey this information to the user and help them gain a sense of their surroundings.

• The current solutions each have a problem. Cheap solutions like sticks with ultrasonic sensors are inaccurate and malfunction in various situations.
• Solutions using LiDar and other light-ray-based sensors are way more expensive and aren't practical for actual usage.
• Field of view of such devices is less than 80 degrees which are way less than what even our eyes can focus on at once.
• Solution- We've made an OpenCV+PyTorch-based algorithm based on the prevalent scale and shift invariant loss algorithm to accurately make a depth map with just a 2d camera feed(for our product the phone's camera). This is then paired with a simple color gradient detection(OpenCV) through which we find the coordinates of the obstacle which then is used to emit sound cues to accurately tell the user where the obstacle is at what distance.
• Compared to other solutions, our product is way cheaper(we're selling an enclosure(3d printed/mold injection)+cloud processing(yearly subscription)). Along with that, it comes with a higher field of view when compared to its competition.