A smart goggles that detects it's surrounding through camera-fed AI model that informs the user with very low or no eyesight,

The problem Drishya solves

Vision plays an important role in the daily life of a person. From getting up to start the day, until ending it, eyes are the most important organs of our body. Unfortunately some people are not able to see this beautiful world. Visually impared people can’t easily recognize obstacles or stairs while using normal blind stick. They struggle with their day to day life and everyday activities.
The blind traveler should depend on any other guide like blind cane , people information , trained dogs, etc. About the 90% of the worlds visually impaired live in developing countries. No safety features are present on the normal blind stick which is mostly used by visually impaired people for their visual aid.

Challenges we ran into

Well we had a few problems while working together with hardware, software and AI Model. We had very much less time to train our own models so we used pre-trained models. But since we had to deploy it on raspberry Pi, we suffered drastically from:

  • low fps due to heavier model,
  • low accuracy due to light models

After trying mobileNet, efficientNet, YOLO, and NanoDet Plus, we settled on the later which provided us with about 5 FPS of Single Shot Detection.