285 million. That's the number of visually impaired people in this world, out of which only around 12% muster the courage to venture outdoors and indulge in regular activities. Constant company is a solution. But the 'constant' cannot be 24*7. And phones are our only 'constant' company. Presenting to you, 'Retina', our venture to eradicate this global impairment of vision, using technology.
What it does
Retina is your personal digital assistant and guide who helps identify objects and read labels. By using any Android device at ones point of view, Retina is capable of being your vision, verbally recreating the scene for you.
How we built it
We build this application using deep learning techniques like Tensor Flow and Google Vision API to identify and classify objects. Built on the Android platform, the application uses Google Web Speech framework to interact with the visually impaired. The trained Tensor Flow Model identifies known objects with greater accuracy. For unknown objects, Retina uses the Google Vision API.
Challenges we ran into
Figuring out the best design for optimal User Experience given that the users are visually impaired was quite a task. Building and training the Tensor Flow model in the limited duration of the Hackathon was a challenge. Being noobs at Android Development, bringing Retina to life was a surreal experience.
Accomplishments that we're proud of
It works. No seriously. IT WORKS!!!
Overcoming the impediments of something as inhibiting as blindness is Herculean. Taking baby steps towards bringing Retina to life and realizing what she is capable of within a day took us by surprise.
Assembling the application together with the little building blocks we learnt through school was EYE opening (pun intended).
What we learned
Gone are the days when mobile phones were assumed to be a part of us. Today, mobile phones have developed to the extent of even substituting a part of us.
What's next for Retina
Retina could be extended to recognizing people (using Facebook API), navigating through streets and spaces (which even Google maps cannot). Retina can also be made to suggest alternate products on the basis of ratings, price and user preference. Additionally, contextual comprehensibility could be the X-Factor in making Retina the ultimate personal assistant.