this project uses convolution kernels to detect shapes in an image that resemble eyes. press scan to track the position of your eyes stay still while it says taking picture, though you can move when it says "scanning" use TurboWarp for way less lag: https://turbowarp.org/751130542 it does not always work. in fact it nearly never works on the first try, and I have only tested it with my camera so i am not sure how well it works with different lighting or backgrounds. for best results: make sure your eyes are open position your head around one foot away from the camera make sure that your hair is not covering your eyes ensure you have relatively good lighting aim your face towards the camera how it works: (I may add comments in the projects to explain better how this works but have not gotten around to doing it yet) it basically uses convolution kernels: https://en.wikipedia.org/wiki/Kernel_(image_processing) what it does is applies a kernel to every pixel on the screen. a kernel is basically a filter that changes to a different value based on surrounding pixels. if you use the correct kernel, it will change brighter if applied to a pixel that is in a line, so you can apply this kernel to every pixel and all lines will become brighter and not line areas will become darker. so with a combonation of these kernels, you can make the pixels that make up eyes brighter. #face #recognition #eye #tracker #face recognition
USE UPDATED VERSION: https://scratch.mit.edu/projects/989454634/
