RGB vs. Cartesian Coordinate Representation¶

To compare the performance of Hyperparameter tuning on the type of representation, HARDy provides easy access to use to control the representation of transformed data in the Convolutional Neural Network (CNN)

The two representations after numerical transformations are shown in the image below: 5da1b0fc8a084adb988975fece30c5ca

The representation can be controlled through plot_format argument in the hardy_main

Using RGB Representation¶

Following code will execute HARDy using RGB representation

run.hardy_main(raw_data_path, tform_config_path, classifier_config_path, batch_size=64,
scale=0.2, num_test_files_class=750, target_size=(100, 100), iterator_mode='arrays',
classifier='tuner', n_threads=1, classes=['class_1', 'class_2', 'class_3'], plot_format='RGBrgb`
project_name='my_project_name')

if the plot_format='RGBrgb', the RGB images will be fed into the neural network

Using Cartesian Representation¶

Any other string value other than RGBrgb will prompt the HARDy to use cartesian representation. The example code will then be:

run.hardy_main(raw_data_path, tform_config_path, classifier_config_path, batch_size=64,
scale=0.2, num_test_files_class=750, target_size=(100, 100), iterator_mode='arrays',
classifier='tuner', n_threads=1, classes=['class_1', 'class_2', 'class_3'], plot_format='cart`
project_name='my_project_name')

By default, the plot_format is kept as ``RGBrgb``