def test_errors_for_bbox_wrong_size(self):
with self.assertRaises(ValueError) as cm:
next(submission_loader.gen_img_pboxes([{
'label_probs': [0.1, 0.1, 0.1, 0.1, 0.2],
'bbox': [12, 14, 55]
}], [list(range(5)), list(range(5))], num_classes=5, img_idx=13, sequence_name='test.json'))
self.assert_contains_indexes(cm, 'test.json', 13, 0)
with self.assertRaises(ValueError) as cm:
next(submission_loader.gen_img_pboxes([{
'label_probs': [0.1, 0.1, 0.1, 0.1, 0.2],
'bbox': [12, 14, 55, 46, 47]
}], [list(range(5)), list(range(5))], num_classes=5, img_idx=13, sequence_name='test.json'))
self.assert_contains_indexes(cm, 'test.json', 13, 0)
def test_errors_for_non_positive_definite_cov(self):
with self.assertRaises(ValueError) as cm:
next(submission_loader.gen_img_pboxes([{
'label_probs': [0.1, 0.1, 0.1, 0.1, 0.2],
'bbox': [12, 14, 55, 46],
'covars': [[[1, 2], [2, 1]], [[1, 0], [0, 1]]]
}], [list(range(5)), list(range(5))], num_classes=5, img_idx=13, sequence_name='test.json'))
self.assert_contains_indexes(cm, 'test.json', 13, 0)
with self.assertRaises(ValueError) as cm:
next(submission_loader.gen_img_pboxes([{
'label_probs': [0.1, 0.1, 0.1, 0.1, 0.2],
'bbox': [12, 14, 55, 46],
'covars': [[[1, 0], [0, 1]], [[1, 2], [2, 1]]]
}], [list(range(5)), list(range(5))], num_classes=5, img_idx=13, sequence_name='test.json'))
self.assert_contains_indexes(cm, 'test.json', 13, 0)
def test_doesnt_normalize_underestimate(self):
det = next(submission_loader.gen_img_pboxes([{
'label_probs': [0.1, 0.1, 0.1, 0.1, 0.2],
'bbox': [12, 14, 55, 46],
'covars': [[[0, 0], [0, 0]], [[0, 0], [0, 0]]]
}], [list(range(5)), list(range(5))], num_classes=5))
expected_list = np.zeros(len(class_list.CLASSES), dtype=np.float32)
expected_list[0:5] = [0.1, 0.1, 0.1, 0.1, 0.2]
self.assertNPEqual(expected_list, det.class_list)
def test_reorders_label_probs(self):
det = next(submission_loader.gen_img_pboxes([{
'label_probs': [0.1, 0.2, 0.3, 0.4],
'bbox': [12, 14, 55, 46],
'covars': [[[0, 0], [0, 0]], [[0, 0], [0, 0]]]
}], [[2, 1, 3], [0, 2, 3]], num_classes=4))
expected_list = np.zeros(len(class_list.CLASSES), dtype=np.float32)
expected_list[2] = 0.1
expected_list[1] = 0.3
expected_list[3] = 0.4
self.assertNPEqual(expected_list, det.class_list)
def test_builds_detection_instances(self):
img_dets = [{
'label_probs': [0.1, 0.1, 0.1, 0.3, 0.4],
'bbox': [12, 14, 55, 46],
'covars': [[[0, 0], [0, 0]], [[0, 0], [0, 0]]]
}, {
'label_probs': [0.8, 0.05, 0.05, 0.05, 0.05],
'bbox': [21, 81, 87, 112],
'covars': [[[15, 5], [5, 21]], [[100, 0], [0, 5]]]
}, {
'label_probs': [0.2, 0.1, 0.3, 0.1, 0.3],
'bbox': [3, 6, 13, 121]
}]
dets = list(submission_loader.gen_img_pboxes(img_dets, [list(range(5)), list(range(5))], num_classes=5))
self.assertEqual(len(img_dets), len(dets))
for idx in range(len(img_dets)):
self.assertIsInstance(dets[idx], data_holders.DetectionInstance)
expected_list = np.zeros(len(class_list.CLASSES), dtype=np.float32)
expected_list[0:5] = img_dets[idx]['label_probs']
self.assertNPEqual(expected_list, dets[idx].class_list)
def test_returns_empty_generator_for_no_detections(self):
gen = submission_loader.gen_img_pboxes({}, [[0], [0]])
self.assertIsInstance(gen, types.GeneratorType)
dets = list(gen)
self.assertEqual(dets, [])
