cuicaihao/mnist_tf_gui

Demonstrate training, testing, and retraining ML/DL models with TensorFlow and Keras with an extra GUI by Qt5

MNIST_TF_GUI README File

Author: Caihao Cui Location: Melbourne, VIC, AU.

Description: This repository demonstrates one python program including training, testing, and retraining ML/DL models with TensorFlow and Keras with an extra GUI by Qt5.

Dependencies: install Anaconda python 3, TensorFlow, Keras and all the package related. TF2.0 works but with few warninig.

Train the model

Use the Train_MNIST_Model.py to build the model, model will be saved at ./models/xxx.xxx.xxx.h5, the training histroy (acc, loss) will be saved as history.pickle.

Test the model

Use the Test_MNIST_Model.py to test the model and compare the training and testing accuracy.

Retrain the model

Use the Retrain_MNIST_Model.py to load the trained model and retrain it with different optimization methods.

Play with GUI

Run GUI.py or MNIST_Window.py to play with the interactive desktop application.

Sample images:

You can change the details by editing the class of MNIST_Window

Sample Images:

Reference

This repository is inspried by the CSDN Blog: https://blog.csdn.net/u011389706/article/details/81460820.