Inspired by the concept of color constancy in human perception, I attempted to make a program that interprets color like humans would. Normally, context illumination is a hinderance to categorizing colors in computer vision. Here, it's the goal.
As shown by the famous dress, visual perception varies by person. Thus, replicating human color constancy is too broad of a task. Because of this, I chose to inform my neural network using probabilities. For example, most people saw the dress as black and blue, so that is what the CNN would report it as.
