def setUp(self):
num_groundtruth_classes = 1
matching_iou_threshold1 = 0.5
matching_iou_threshold2 = 0.1
self.eval1 = per_image_evaluation_rbox.PerImageEvaluationRbox(
num_groundtruth_classes, matching_iou_threshold1)
self.eval2 = per_image_evaluation_rbox.PerImageEvaluationRbox(
num_groundtruth_classes, matching_iou_threshold2)
self.detected_boxes = np.array(
[[0, 0, 1, 1, 0.0], [0, 0, 2, 2, 0.0], [0, 0, 3, 3, 0.0]],
dtype=np.float32)
self.detected_scores = np.array([0.6, 0.8, 0.5], dtype=np.float32)
def test_tp_fp(self):
num_groundtruth_classes = 3
matching_iou_threshold = 0.5
eval1 = per_image_evaluation_rbox.PerImageEvaluationRbox(
num_groundtruth_classes, matching_iou_threshold)
detected_boxes = np.array(
[[0, 0, 1, 1, 0.0], [0, 0, 2, 2, 0.0], [0, 0, 3, 3, 0.0]],
dtype=np.float32)
detected_scores = np.array([0.8, 0.8, 0.8], dtype=np.float32)
detected_class_labels = np.array([0, 1, 2], dtype=int)
groundtruth_boxes = np.array(
[[0, 0, 1, 1, 0.0], [0, 0, 3.5, 3.5, 0.0]], dtype=np.float32)
groundtruth_class_labels = np.array([0, 2], dtype=int)
groundtruth_groundtruth_is_difficult_list = np.zeros(2,
dtype=np.float32)
scores, tp_fp_labels, _ = eval1.compute_object_detection_metrics(
detected_boxes, detected_scores, detected_class_labels,
groundtruth_boxes, groundtruth_class_labels,
groundtruth_groundtruth_is_difficult_list)
expected_scores = [np.array([0.8], dtype=np.float32)] * 3
expected_tp_fp_labels = [
np.array([True]),
np.array([False]),
np.array([True])
]
for i in range(len(expected_scores)):
self.assertTrue(np.allclose(expected_scores[i], scores[i]))
self.assertTrue(
np.array_equal(expected_tp_fp_labels[i], tp_fp_labels[i]))
def test_compute_corloc_with_normal_iou_threshold(self):
num_groundtruth_classes = 3
matching_iou_threshold = 0.5
eval1 = per_image_evaluation_rbox.PerImageEvaluationRbox(
num_groundtruth_classes, matching_iou_threshold)
detected_boxes = np.array([[0, 0, 1, 1, 0], [0, 0, 2, 2, 0],
[0, 0, 3, 3, 0], [0, 0, 5, 5, 0]],
dtype=np.float32)
detected_scores = np.array([0.9, 0.9, 0.1, 0.9], dtype=np.float32)
detected_class_labels = np.array([0, 1, 0, 2], dtype=int)
groundtruth_boxes = np.array(
[[0, 0, 1, 1, 0], [0, 0, 3, 3, 0], [0, 0, 6, 6, 0]],
dtype=np.float32)
groundtruth_class_labels = np.array([0, 0, 2], dtype=int)
is_class_correctly_detected_in_image = eval1._compute_cor_loc(
detected_boxes, detected_scores, detected_class_labels,
groundtruth_boxes, groundtruth_class_labels)
expected_result = np.array([1, 0, 1], dtype=int)
self.assertTrue(
np.array_equal(expected_result,
is_class_correctly_detected_in_image))
def __init__(self,
num_groundtruth_classes,
matching_iou_threshold=0.5,
nms_iou_threshold=1.0,
nms_max_output_boxes=10000,
box_type='box',
is_save_detection_results=False):
self.box_type = box_type
if box_type == 'rbox':
self.per_image_eval = per_image_evaluation_rbox.PerImageEvaluationRbox(
num_groundtruth_classes, matching_iou_threshold,
is_save_detection_results)
else: # box_type == 'box'
self.per_image_eval = per_image_evaluation.PerImageEvaluation(
num_groundtruth_classes, matching_iou_threshold,
nms_iou_threshold, nms_max_output_boxes)
self.num_class = num_groundtruth_classes
self.groundtruth_boxes = {}
self.groundtruth_class_labels = {}
self.groundtruth_is_difficult_list = {}
self.num_gt_instances_per_class = np.zeros(self.num_class, dtype=int)
self.num_gt_imgs_per_class = np.zeros(self.num_class, dtype=int)
self.detection_keys = set()
self.scores_per_class = [[] for _ in range(self.num_class)]
self.tp_fp_labels_per_class = [[] for _ in range(self.num_class)]
self.num_images_correctly_detected_per_class = np.zeros(self.num_class)
self.average_precision_per_class = np.empty(self.num_class,
dtype=float)
self.average_precision_per_class.fill(np.nan)
self.precisions_per_class = []
self.recalls_per_class = []
self.corloc_per_class = np.ones(self.num_class, dtype=float)
self.is_save_detection_results = is_save_detection_results
self.detection_results = [] if is_save_detection_results else None