Artificial Intelligence Digital Assistant For Visually Impaired People

  • Assistant Professor et al.

Abstract

The World Health Organization states that, in 2017, 253 million people face vision impairment: 36
million are blind and 217 million have moderate to severe problem for vision impairment .In India
currently around 12 million blind people are there, which puts India home to one-third of the world’s
blind population which is not a good number. We want to assist visually impaired people living in India
by using the camera of an Android phone to capture images, identify them and provide information
related to its which would improve their day-to-day routine. The application includes currency detection,
product recognition using barcode and text reader. The application is trained using Convolutional
Neural Network (CNN) to recognize the currency and by applying transfer learning over the model VGG
16 and Keras framework (Python). Barcode is detected using Mobile Vision API provided by Google and
the information about the product is fetched using other public API’s. Similarly, Mobile Vision API is
used to read text from an image clicked by a user. The app is gesture-based, and the detected results are
converted to speech, thus making it easier for the unsighted person to use.

Published
2020-05-01