def test_complains_on_zero_remaining(self):
num_images_per_class = np.array([10, 10, 10, 10, 10])
num_remaining_per_class = np.array([5, 0, 5, 5, 5])
support_set_size = 5
with self.assertRaises(ValueError):
sampling.sample_num_support_per_class(
num_images_per_class,
num_remaining_per_class,
support_set_size,
min_log_weight=test_utils.MIN_LOG_WEIGHT,
max_log_weight=test_utils.MAX_LOG_WEIGHT)
def test_at_least_one_example_per_class(self):
num_images_per_class = np.array([10, 10, 10, 10, 10])
num_remaining_per_class = np.array([5, 5, 5, 5, 5])
support_set_size = 5
for _ in range(10):
num_support_per_class = sampling.sample_num_support_per_class(
num_images_per_class,
num_remaining_per_class,
support_set_size,
min_log_weight=test_utils.MIN_LOG_WEIGHT,
max_log_weight=test_utils.MAX_LOG_WEIGHT)
self.assertTrue((num_support_per_class > 0).any())
def test_support_set_size_respected(self):
num_images_per_class = np.array([50, 40, 30, 20])
num_remaining_per_class = np.array([40, 30, 20, 10])
support_set_size = 50
for _ in range(10):
num_support_per_class = sampling.sample_num_support_per_class(
num_images_per_class,
num_remaining_per_class,
support_set_size,
min_log_weight=test_utils.MIN_LOG_WEIGHT,
max_log_weight=test_utils.MAX_LOG_WEIGHT)
self.assertLessEqual(num_support_per_class.sum(), support_set_size)
