Programmatically, how are you judging the similarity between the user's painting and the model painting?
It works off taking a screenshot of the screen at the end of the timer, the pixel colours are comapred between the reference image and the nail with the relevant scaling factor. Each pixel in the nail is checked and the percentage accuracy is multiplied by the difficulty of the design. Hope that helped :) If you want more info, the code is very messy but I could upload it to github if needed.