https://www.bilibili.com/video/av88086690/
Blog：https://blog.valderfield.com/archives/17/

A handwriting number recognizer based on TensorFlow.

MNIST is the "Hello, World" in machine learning. I add an drawingboard, although it doesn't work fluently.

Run demo1.py in the directory within python3, TensorFlow 2 environment. Then you can write number on the drawingboard. Every time the left button of your mouse up, the program will recognize the image on the board as a number, and print the result in the terminal. Press ESC to quit the program, and press C to clear the drawingboard.

demo1.py imports drawingboard.py(the drawingboard UI) and ml.py(which includes the machine learning model using the trained model file "my_model_eproch_5.h5" or "my_model.h5").

If you want to train the model by yourself, you can run train.py on your own computer or on the CoLab by Google. You may learn how to copy the trained model to your Google Drive if you choose the last one.

img2arr.py is a program that converts an image into an array which is the input of the neural network. It is intergreted into ml.py as a method.