Regression training

operators.ml3d.regression_training(client, in_folder_points='data_train/points', in_folder_gt='data_train/gt', out_model_path='parameters_model', voxel_size=0.02, zero_centering='True', point_names='X,Y,Z', feature_names='ones', learning_rate=3e-06, weight_decay=1e-09, num_epochs=500, batch_size=2, save_after_epochs=100, network_type='FCNN128', criterion_type='L1Mean', num_classes=1, instance_type='x2large')

regression_training( client,

in_folder_points=’data_train/points’,

in_folder_gt=’data_train/gt’,

out_model_path=’parameters_model’,

voxel_size=0.02,

zero_centering=’True’,

point_names=’X,Y,Z’,

feature_names=’ones’,

learning_rate=3e-6,

weight_decay=1e-9,

num_epochs=500,

batch_size=2,

save_after_epochs=100,

network_type=’FCNN128’,

criterion_type=’L1Mean’,

num_classes=1,

instance_type=’x2large’ )

Parameters:

• in_folder_points – input directory with training data

• in_folder_gt – input directory with training data

• out_model_path – model path

• voxel_size – voxel size

• zero_centering – zero centering

• point_names – point names

• feature_names – feature names

• learning_rate – learning rate

• weight_decay – regularization decay

• num_epochs – number of epochs

• batch_size – batch size for training

• save_after_epochs – save after epochs

• network_type – model type of backbone network

• criterion_type – model type of criterion

• num_classes – number of classes

• instance_type – type of cloud instance used for processing