def load_dataset(sourcefolder, split_factor):
dataset = Crosswalk_dataset.from_sourcefolder(sourcefolder)
dataset.read_samples()
dataset.load_images()
(train_set, test_set) = dataset.split_train_test(split_factor)
_print_dataset_details(dataset, train_set, test_set)
return (train_set.to_input_response(), test_set.to_input_response())
def test_dataset_split(self):
dataset_path = self.get_test_dataset_path()
factor = 0.6
set = Crosswalk_dataset.from_sourcefolder(dataset_path)
set.read_samples()
(train_set, test_set) = set.split_train_test(factor)
self.assertEquals(len(train_set.samples_shuffled), 6)
self.assertEquals(len(test_set.samples_shuffled), 4)
def testDataset_read_samples(self):
dataset_path = self.get_test_dataset_path()
set = Crosswalk_dataset.from_sourcefolder(dataset_path)
set.read_samples()
nb_samples = 10
self.assertEquals(len(set.samples_shuffled), nb_samples)
self.assertGreater(len(set.samples_crosswalk.samples), 0)
self.assertGreater(len(set.samples_nocrosswalk.samples), 0)
def test_dataset_split_nodouble(self):
dataset_path = self.get_test_dataset_path()
factor = 0.5
set = Crosswalk_dataset.from_sourcefolder(dataset_path)
set.read_samples()
(train_set, test_set) = set.split_train_test(factor)
for test_sample in test_set.samples_shuffled:
for train_sample in train_set.samples_shuffled:
self.assertNotEqual(test_sample.filepath, train_sample.filepath)
def test_dataset_to_input_response(self):
dataset_path = self.get_test_dataset_path()
set = Crosswalk_dataset.from_sourcefolder(dataset_path)
set.read_samples()
set.load_images()
(inputs, responses) = set.to_input_response()
should_input_shape = (10, 3, 50, 50)
should_response_shape = (10,2)
self.assertEquals(inputs.shape, should_input_shape)
self.assertEquals(responses.shape, should_response_shape)
def test_dataset_load_images(self):
dataset_path = self.get_test_dataset_path()
set = Crosswalk_dataset.from_sourcefolder(dataset_path)
set.read_samples()
set.load_images()
self.assertIsNotNone(set.samples_shuffled[0].pil_image)
