Skip to content

Latest commit

 

History

History
77 lines (55 loc) · 3.23 KB

README.md

File metadata and controls

77 lines (55 loc) · 3.23 KB
Raw

Currently the project is work in progress. See an older version of the website & PWA here.

Orama Visual Assistant banner

Orama Visual Assistant

Pitch video
Story behind the project
Project time line
WIP / For future
License

Orama Visual Assistant (from Greek: Όραμα, meaning "vision") is an app for visually impaired people that announces objects detected using user's phone camera.

Our goal is to help visually impaired people to become more independent.

Pitch video


Link to the pitch video on YouTube

Story behind the project

We are the Nurmukhambetov family. There are seven of us in the family: mom, dad, four sons and a daughter. Three of the sons seen in the photograph developed this project.

“Our youngest daughter, Sofia, drove us to create this project. Sofia, was born prematurely, from birth we fought for her life, they tried as much as possible to save her eyesight. In consequence of 12 operations, we were able to save Sofia's light perception. But this is not enough, in order to live a full-fledged life on her own, she needs outside help.”

‒ Damira Nurmukhambetova

Project time line

Nov 2019
We began brainstorming on the ideas.
Dec 2019
First demo made using KNN algorithm on images.
Apr 2020
Classification MobileNetv2 model trained with transfer learning (Teachable Machine).
May 2020
Third place finalists at HackDay 2020. First funding of the project.
Jul 2020
Finalists across the Aisa continent at Technovation Families.
Jan 2022
Change from image classification to object detection (SSD Mobile Netv2). Change from naming everything to search for an object task.
OCR functionality with Tesseract. It is there, but it has bad performance.
Mar 2022
Began working on a react native version of the app.

WIP / For future

  • Improve OCR reader.
  • Change to React Native.
  • Transfer to cloud processing rather than on-device.
  • Currency classification. Name what banknote that is seen in the camera.
  • Facial recognition. Each user will have their own pool of people saved in the app for future facial recognition tasks.
  • Image description. Describe what's in the image.
  • 'Open with OramaVA' on images to do OCR/ facial recognition/ image description.

License

All of the codebase is MIT Licensed unless otherwise stated.