A visual aid glasses for visually impaired or blind people
- 0 Collaborators
In a nutshell, I want to make glasses for the visually impaired which can help them "seeing" things by hearing. ...learn more
Project status: Under Development
Internet of Things, Artificial Intelligence
Intel Technologies
AI DevCloud / Xeon,
Movidius NCS
Overview / Usage
In the prototype version, I am proposing to include 3 main features for the user:
-
What is in front of the user? This is probably the best feature of the glass as It will help them directly see what the glasses are seeing.
This will be solved by any simple Image classification deep learning framework or some advanced versions of a classification framework using transfer learning just as in the paper, which I coauthored - the rice classification model using transfer learning. -
The direction of the things of which are in the way. The user can know from which direction a person or an animal is coming or where is a chair etc.
This can be done using image segmentation. Which can be done using a pretty standard deep learning framework of image translation GANs. -
Image Captioning. Now, this feature is more advanced and may not be handled by the Neural Compute Stick. But, if I can incorporate this feature also in the glasses then the product will have a personal touch.
This can be done by image captioning models using a mixture of CNN and RNN and can be improved using augmentation methods which I have proposed in the Image captioning project.
If this project gets funding then it would be a great thing for the visually impaireds.
Image source: https://www.google.co.in/url?sa=i&rct=j&q=&esrc=s&source=images&cd=&cad=rja&uact=8&ved=0ahUKEwjH-qXTwZPXAhVMK48KHTqNC3AQjRwIBw&url=https%3A%2F%2Fwww.trendhunter.com%2Ftrends%2Fnavigation-blind-sunglasses&psig=AOvVaw1OrqFx63CW0RPwSIqj244K&ust=1509286797322541
https://www.google.co.in/url?sa=i&rct=j&q=&esrc=s&source=images&cd=&cad=rja&uact=8&ved=0ahUKEwjck_K7wpPXAhXJsY8KHWKMBxYQjRwIBw&url=https%3A%2F%2Fdevblogs.nvidia.com%2Fparallelforall%2Fimage-segmentation-using-digits-5%2F&psig=AOvVaw1D2WBYTN-zczWSvfbsr1U9&ust=1509287028458170
Methodology / Approach
CNN, RNN, GAN, Pix2Pix, Image captionioning.
Technologies Used
Intel DevCloud, Intel Movidius Neural Compute Stick.