akshit17/colorblindaid

A application for people suffering from color-blindness to help them see their surrounding colors real-time through webcam/mobile camera.

colorblindaid's Issues

Structure of Application

Since the goal is to create a real-time stream for color-blind people to see the true color of frames presented through their phone camera/ webcam, a server-based application is preferably expected to use. But this leads to the problem of Requirement of good internet from the user side and Network Latency and real-time stream might increase load and increase response time of API.

Another alternative can be running the code on a local device itself , advantages of this are that it won't require the user to have an internet connection but would increase the load on the device and real-time stream would require the user to have a large empty memory.

Algorithm of program

Color-Blindness(CB) can differ from person to person . The most common ones are Red-Green CB , Blue-yellow CB , Complete CB.
Types Of Color-Blindness

Considering that we want to output a real-time stream that rewrites/masks the color of an object in the frame in such a way that the colorblind user is able to see the actual color of the object, we might require a calibration program to test the type of color blindness of user or allow him to input by himself if he knows it beforehand.

Taking in account what type of CB the user has , we make changes to the frame accordingly.
However simply changing every color pixel the user is not able to see can lead to a distorted image.

Intuitive thinking concludes deep learning approach for creating the color correction filter might be required.
Colorspace to use for Correction filter is unclear. Simply using RGB, HSV/HSL might not give accurate results. Why use LMS color space?.
Algorithm to use the LMS colorspace to visualize the view of a colorblind and color it accordingly is unclear.