A Mobile Application CODEC (Color Detection) for Color-blind People using KNN

Color blindness is a disease generally acquired since birth and cannot be cured. Over time, the limitations in seeing colors will further hinder color-blind people. The research aims to develop an application to help color-blind people distinguish between these colors when pointing an object. The application implements the K-Nearest Neighbors (KNN) algorithm to classify color as an Android application using the React Native framework. We collected images as the dataset using mobile phones. The dataset has 500 images (50 images representing each of ten colors in RGB values). This dataset already provides the needed information, namely the features (RGB values) and the labels (color’s name). We then processed the dataset using the KNN model available on the REST API. This machine learning-based color detection mobile app can detect ten different colors we often see daily: red, orange, yellow, green, blue, purple, brown, black, white, and gray. The color detection system was tested with three different light intensities. The average accuracy of the detection results for the ten colors is 92% (high light intensity), 97% (medium light), and 76% (lowest light). It shows that the application performs well in detecting ten colors in an object and can help color-blind people distinguish between them in everyday life.