I have no memory for faces. This can make remembering the names of new people rather difficult. For the blind or those at conferences, or if you just generally have a poor memory like me, Whispr is a pair of glasses utilizing image recognition and machine learning to recognize your friends, find their internet presences, and provide a little extra info for the curious. Whispr is a unique personal assistance device meant to aid in the speedy recognition of those around you and help you socialize smoothly. Whispr is a cloud solution for blindness and prosopagnosia.
What it does
Whispr is a camera, currently integrated into a pair of sunglasses, connected to a portable raspberry pi that can take and process images of people around you and let you know who they are through a pair of headphones connected to the pi, or through the watch app we built.
How we built it
The pi is running a python server to take requests for new images and send the images to our server for processing and recognition, then receive a voice files from AWS Polly and Rekognition and read them back to the user. We also built a Pebble watch app to trigger new images and display a list of the people you've seen recently.
Challenges we ran into
Facebook has no api for their facial recognition. AWS refused to approve our account for around 12 hours.
Accomplishments that we're proud of
Built custom apis for facebook and linkedin.
What we learned
A fair bit about using AWS to build microservices within your project. How to create a fake facebook api and emulate refresh tokens.
What's next for Whispr
Integration of Whispr hardware in a variety of other functional and fashionable ways to recognize your surroundings, such as a ties and handbags. More robust representations of information, reduced roundtrip on new image requests. All this will come together to make Whispr better than Google Glass.