# # Normalise data to zero mean and unit variance.
# scaler = StandardScaler()
# normalized_hinselmann_data = scaler.fit_transform(hinselmann_data)
# normalized_green_data = scaler.fit_transform(green_data)
# normalized_schiller_data = scaler.fit_transform(schiller_data)
# Define all possible {source, target} pairs from list of modalities.
modalities = ['h', 'g', 's']
tasks = permutations(modalities, r=2)
# Loop through each task.
for (source, target) in tasks:
# Define file path for results.
dataset_str = 'colposcopy_{}_{}'.format(source, target)
dp_gp_lvm_results_file = RESULTS_FILE_NAME.format(model='dp_gp_lvm',
dataset=dataset_str)
# Visualise results if they exist.
if isfile(dp_gp_lvm_results_file):
results = np.load(dp_gp_lvm_results_file)
ground_truth = results['y_test_unobserved'][:, :-1]
predicted_mean_assessments = results['predicted_mean'][:, :-1]
plot.imshow(ground_truth,
interpolation='nearest',
aspect='auto',
extent=(0, ground_truth.shape[1],
ground_truth.shape[0], 0),
origin='upper')
plot.colorbar()
from src.utils.constants import ResultKeys, RESULTS_FILE_NAME, PLOTS_PATH
import src.visualisation.plotters as vis
import matplotlib.pyplot as plot
import numpy as np
from os.path import isfile
if __name__ == '__main__':
# Define booleans. TODO: Allow them to be set from command line.
save_plots = False
show_plots = True
dataset_str = 'cmu_walking_normal_swapped'
bgplvm_results_file = RESULTS_FILE_NAME.format(model='bgplvm',
dataset=dataset_str)
# 93 dims per view.
mrd_results_file = RESULTS_FILE_NAME.format(model='mrd',
dataset=dataset_str)
# 3 dims per view so keep 3d points together.
mrd_fully_independent_results_file = RESULTS_FILE_NAME.format(
model='mrd_fully_independent', dataset=dataset_str)
# Keep 3d points together so mask of 3.
gpdp_results_file = RESULTS_FILE_NAME.format(model='dp_gp_lvm',
dataset=dataset_str)
# Keep skeletons together so mask of 93.
gpdp_mask_results_file = RESULTS_FILE_NAME.format(
model='dp_gp_lvm_mask_93', dataset=dataset_str)
normal_swapped_motions[::4]) # subsample frames.
num_samples, num_output_dimensions = y_train.shape
# Print info.
print('\nCMU Walking 35 with Normal and Swapped Legs Motion:')
print(' Total number of observations (N): {}'.format(num_samples))
print(' Total number of output dimensions (D): {}'.format(
num_output_dimensions))
print(' Total number of inducing points (M): {}'.format(
num_inducing_points))
print(' Total number of latent dimensions (Q): {}'.format(
num_latent_dimensions))
# Define file path for results.
dataset_str = 'cmu_walking_normal_swapped'
bgplvm_results_file = RESULTS_FILE_NAME.format(model='bgplvm',
dataset=dataset_str)
mrd_results_file = RESULTS_FILE_NAME.format(
model='mrd', dataset=dataset_str) # 93 dims per view.
# 3 dims per view so keep 3d points together.
mrd_fully_independent_results_file = RESULTS_FILE_NAME.format(
model='mrd_fully_independent', dataset=dataset_str)
gpdp_results_file = RESULTS_FILE_NAME.format(
model='dp_gp_lvm', dataset=dataset_str) # Keep 3d points together.
gpdp_mask_results_file = RESULTS_FILE_NAME.format(
model='dp_gp_lvm_mask_93', dataset=dataset_str)
# Define instance of necessary model.
if not isfile(gpdp_results_file):
# Reset default graph before building new model graph. This speeds up script.
tf.reset_default_graph()
np.random.seed(1) # Random seed.
:return:
"""
file_path = skin_cancer_mnist_path + 'hmnist_8_8_L.csv'
observed_data, diagnosis_labels = read_hmnist_csv(file_path)
assert observed_data.shape[
1] == 64, 'Number of pixels does not match expected number of 64.'
return observed_data, diagnosis_labels
if __name__ == '__main__':
# Define file path for results.
bgplvm_results_file = RESULTS_FILE_NAME.format(model='bgplvm',
dataset='skin_cancer_mnist')
mrd_results_file = RESULTS_FILE_NAME.format(model='mrd',
dataset='skin_cancer_mnist')
mrd_fully_independent_results_file = RESULTS_FILE_NAME.format(
model='mrd_fully_independent', dataset='skin_cancer_mnist')
gpdp_results_file = RESULTS_FILE_NAME.format(model='dp_gp_lvm',
dataset='skin_cancer_mnist')
gpdp_mask_results_file = RESULTS_FILE_NAME.format(
model='dp_gp_lvm_mask_64', dataset='skin_cancer_mnist')
# Choose what model we are looking at.
results_file = bgplvm_results_file
if isfile(results_file):
# Read results.
results = np.load(results_file)