def perform_loocv_classifiers(args):
# Instantiate Classifier
classifier = Classifier(classifier=None,
no_clusters=args.number_cluster,
no_samples=None)
classifier.OsName = platform.system()
print('Operating System: ', classifier.OsName)
print('LOOCV Classification')
classifier.test_name = args.test_name
classifier.aggregateVideoFeatures = False
if args.use_train_test_val:
classifier.datasets = ['training', 'validation', 'test']
classifier.base_path = args.base_path
classifier.label_path = args.label_path
# Send e-mail Process Start
send_email_start(classifier)
# train the model
classifier.trainModelFV_LOOCV_Classifiers(extension=args.features_file_filter)
def perform_loocv_fusion(args):
classifier = Classifier(classifier=SVC(kernel='linear'),
no_clusters=args.number_cluster,
no_samples=None)
classifier.OsName = platform.system()
print('Operating System: ', classifier.OsName)
print('LOOCV Fusion')
classifier.test_name = args.test_name
classifier.aggregateVideoFeatures = False
if args.use_train_test_val:
classifier.datasets = ['training', 'validation', 'test']
classifier.base_path = args.base_path
classifier.base_path2 = args.base_path2
classifier.features_file_filter = args.features_file_filter
classifier.label_path = args.label_path
# Send e-mail Process Start
send_email_start(classifier, True)
# train the model
classifier.trainModelFV_LOOCV_Fusion()
def main():
OsName = platform.system()
print('Operating System: ', OsName)
parser = argparse.ArgumentParser(
description="Compute trajectory features from OpenPose points to Human Action Recognition"
)
parser.add_argument("--features_file_filter", type=str,
default='*.*',
help="Filter for features files")
parser.add_argument("--base_path", type=str, required=True,
help="Base features path.")
parser.add_argument("--base_path2", type=str,
help="Base features path2 for features fusion.")
parser.add_argument("--output_features", type=str, required=True,
help="Output features file name.")
parser.add_argument("--output_image", type=str, required=True,
help="Output image path.")
parser.add_argument("--label_path", type=str,
help="Labels path.")
parser.add_argument("--number_cluster", type=int,
default=20,
help="Number of cluster for FV.")
parser.add_argument("--number_classes", type=int,
default=6,
help="Number of classes of dataset.")
parser.add_argument("--mark_size", type=int,
default=100,
help="Size of the points.")
parser.add_argument("--alpha", type=float,
default=0.6,
help="Proportional opacity of the points.")
parser.add_argument("--use_train_test_val", type=int,
default=0,
help="True if dataset is divides into train/test/validation.")
args = parser.parse_args()
print(args)
file_name = args.output_features
if os.path.isfile(file_name):
df = pd.read_csv(file_name, index_col=0)
print(df.head())
plotDataVisualization(df.drop('label', axis=1).values, df.label.values, 'load_full', args)
else:
classifier = Classifier(no_clusters=args.number_cluster)
if args.use_train_test_val:
classifier.datasets = ['training', 'validation', 'test']
classifier.base_path = args.base_path
classifier.base_path2 = args.base_path2
classifier.label_path = args.label_path
X, y = classifier.build_FV_Features()
df = pd.DataFrame(X, y)
df = df.reset_index()
df = df.rename(columns={df.columns[0]: "label"})
print(df.head())
df.to_csv(file_name)
plotDataVisualization(df.drop('label', axis=1).values, df.label.values, 'dataframe_full', args)
plotDataVisualization(X, y, 'ndarray_full', args)
