Check out model1.py and model2.py to see how the models were trained. For both of the different CNN designs there is a pre trained model. Model1.h5 was trained in model.py with a accuracy of about 84%. Model2.h5 was trained in model2.py and achieved an accuracy of about 92% In prediction.py you can use the pre trained models to classify your own images.
In order to run the python files you need to setup the followning modules:
- tensorflow (check out the installation guide https://www.tensorflow.org/install)
- numpy (https://pypi.org/project/numpy/)
- matplotlib (https://pypi.org/project/matplotlib/)