1 of 9

ASSISTIVE �TECHNOLOGY FOR THE VISUALLY IMPAIRED

AN A.I-POWERED APPLICATION

2 of 9

PROBLEM STATEMENT

  • Currency is used almost everywhere to facilitate business.
  • Most developing countries in the world especially in Africa a tangible note is used predominantly.
  • Globally, the visually impaired population is estimated to be 285 million, of whom 39 million are vision loss (W.H.O.).
  • Business transactions by the blind have been extremely difficult. For instance, In shopping center, super markets, bus stations, every where.

3 of 9

PREVIOUS WORKS

  • There have been few researches focusing on Counterfeits detection.
  • There haven’t been works that provide a solution to the visually impaired people in identifying currencies Especially in Africa.

4 of 9

PROPOSED SOLUTION

  • An android and IOS compatible mobile application that recognizes the Banknote different countries.
  • The application has a voice integrated with it. It provides a sound when it is opened and tells the currency hold in hand in Amharic, a working language of Ethiopia.
  • The application is currently developed for the newly released Ethiopian Banknote (5,10,50,100,200).

5 of 9

METHODOLOGY

  • A Convolutional Neural Network (CNN) is used to develop the model. And TensorFlow lite is used to support a mobile application.
  • A total dataset of 2060 images are used for developing and testing including the Unknown class.
  • The model achieved 98.9% accuracy.

6 of 9

CONCLUSION

  • Currently, we are working with the Ethiopian National Association of the Blind (ENAB) to test the usability of the application.
  • Widespread adoption of this technology will greatly enhance the ability of the visually impaired to transact business independently.

7 of 9

FUTURE WORKS

  • We are working to integrate more languages of Ethiopia to the application and reach more people.
  • Work on other African countries to support more visually impaired people.
  • Support for more objects or tools that are routinely used by the visually impaired.

8 of 9

Questions?

9 of 9

THANK YOU!!