def test_sample_dataset_10(self): _, _, sample_train_dataset, sample_valid_dataset = make_train_and_valid_loaders( 1, max_images=100) items = sample_dataset(sample_train_dataset, 10) self.assertEqual(items.shape[0], 10)
def test_unequal_split(self): _, _, sample_train_dataset, sample_valid_dataset = make_train_and_valid_loaders( 2, max_images=100, proportion_faces=0.6) labels = [i[1][1] for i in list(enumerate(sample_train_dataset))] face_labels = [i for i in labels if i == 1] nonface_labels = [i for i in labels if i == 0] self.assertGreater(len(face_labels), len(nonface_labels))
def test_5050_split(self): _, _, sample_train_dataset, sample_valid_dataset = make_train_and_valid_loaders( 2, max_images=100) labels = [i[1][1] for i in list(enumerate(sample_train_dataset))] face_labels = [i for i in labels if i == 1] nonface_labels = [i for i in labels if i == 0] self.assertEqual(len(face_labels), len(nonface_labels))
def test_draw_10_samples_without_max_image(self): _, _, sample_train_dataset, sample_valid_dataset = make_train_and_valid_loaders( 1) idxs_train = np.random.choice(len(sample_train_dataset), 10) idxs_valid = np.random.choice(len(sample_valid_dataset), 10) for idx in idxs_train: sample_train_dataset[idx] for idx in idxs_valid: sample_valid_dataset[idx] sample_train_dataset[15600] sample_valid_dataset[5600]
def test_variety_in_training(self): _, _, sample_train_dataset, sample_valid_dataset = make_train_and_valid_loaders( 1, max_images=100) counter = Counter() for idx, _ in enumerate(sample_train_dataset): item, label, idx = sample_train_dataset[idx] if label == 0: counter['nonface'] += 1 else: counter['face'] += 1 self.assertGreaterEqual(counter['face'], 30) self.assertGreaterEqual(counter['nonface'], 30)
def test_draws_max_100_total_samples(self): _, _, sample_train_dataset, sample_valid_dataset = make_train_and_valid_loaders( 2, max_images=100) self.assertEqual( len(sample_train_dataset) + len(sample_valid_dataset), 100)
def sample_idxs_from_valid_dataset(self): sample_train_loader, sample_valid_loader, sample_train_dataset, sample_valid_dataset = make_train_and_valid_loaders( 1, max_images=100) subsample = sample_idxs_from_sub_dataset([100, 200, 700], sample_valid_loader, 1) return subsample