This is a digit detector I made in java from scratch. It uses a neural network, autocoder and a cluster to detect what digit is in a 25 by 25 pixel image of a digit. It starts of by training the data in an neural network of 625 as the input layer then 100,10,100 as the hidden layers and finally 625 is the output layer. The 625 is because the image is 25 by 25 pixels which is 625. The neural network acts as an auto encoder to take an image converts it to the middle of the network which has 10 doubles the decodes this back to an image. The program works by taking that middle layer of 10 doubles and check to see what is the closest 3 to it. I then finds what number that those were and finds out what number itself is. Its 50% accurate but I made it all pretty much from scratch