Python Forum

Hi!

So i've been training an number classification program with the Mnist dataset, and i've also tested it with Mnist:s "test set" and got it to work pretty good (95% accuracy). But my goal is to create a program where you can paint an image yourself and then let the program classify the specific number. The problem is just that i do not know how to create an image with similar values to a Mnist image. If someone is familiar with this process (paint an image, and make my program under stand the image) please help me!

If you want to generate images, I would suggest to look at deep learning techniques, e.g. GAN (generative adversarial networks) and/or VAE (variational autoencoders). You can train a GAN model on MNIST images and use the model to generate similar (as in MNIST) images.