Convolutional neural network to classify images of the PneumoniaMNIST dataset. For educational purposes. Code with docstring and type hints.
The model hyperparameters are tuned with KerasTuner using the Hyperband optimization algorithms. The hyperparameters are:
- Learning rate
- Number of convolutional layers
- Number of convolutional filters
- Dropout
- Number of neurons in the dense layer
A utility function is also available for plotting the feature maps (random channels):
- Downloads last month
- 0
Evaluation results
- Accuracyself-reported0.966