Bulding a few convolutional neural networks to classify MNIST dataset.
- Prepare data to feed into a model.
- Choose hidden layer for the model.
- Diplay result using accuracy plot, confusion matrix, and classification report.
- Check summary of a model to see number of parameters on each hidden layer.
Training the final CNN that classifies the maximum out of the test set.
- Find how many entries were missclasified
- Diplay missclassified entries
- Look inside of the model's layers: see how filters from different layers applied on an image
- Print filters' weights, vizualize filter's as images
- Predict pesonal handwritten numbers using trained model from 0 to 9