def setUp(self):
num_groundtruth_classes = 1
matching_iou_threshold_high_iou = 0.5
matching_iou_threshold_low_iou = 0.1
nms_iou_threshold = 1.0
nms_max_output_boxes = 10000
self.eval_high_iou = per_image_evaluation.PerImageEvaluation(
num_groundtruth_classes, matching_iou_threshold_high_iou,
nms_iou_threshold, nms_max_output_boxes)
self.eval_low_iou = per_image_evaluation.PerImageEvaluation(
num_groundtruth_classes, matching_iou_threshold_low_iou,
nms_iou_threshold, nms_max_output_boxes)
self.detected_boxes = np.array(
[[0, 0, 1, 1], [0, 0, 2, 2], [0, 0, 3, 3]], dtype=float)
self.detected_scores = np.array([0.6, 0.8, 0.5], dtype=float)
detected_masks_0 = np.array([[0, 1, 1, 0], [0, 0, 1, 0], [0, 0, 0, 0]],
dtype=np.uint8)
detected_masks_1 = np.array([[1, 0, 0, 0], [1, 1, 0, 0], [0, 0, 0, 0]],
dtype=np.uint8)
detected_masks_2 = np.array([[0, 0, 0, 0], [0, 1, 1, 0], [0, 1, 0, 0]],
dtype=np.uint8)
self.detected_masks = np.stack(
[detected_masks_0, detected_masks_1, detected_masks_2], axis=0)
def __init__(self,
num_groundtruth_classes,
matching_iou_threshold=0.5,
nms_iou_threshold=1.0,
nms_max_output_boxes=10000,
use_weighted_mean_ap=False,
label_id_offset=0,
group_of_weight=0.0):
if num_groundtruth_classes < 1:
raise ValueError(
'Need at least 1 groundtruth class for evaluation.')
self.per_image_eval = per_image_evaluation.PerImageEvaluation(
num_groundtruth_classes=num_groundtruth_classes,
matching_iou_threshold=matching_iou_threshold,
nms_iou_threshold=nms_iou_threshold,
nms_max_output_boxes=nms_max_output_boxes,
group_of_weight=group_of_weight)
self.group_of_weight = group_of_weight
self.num_class = num_groundtruth_classes
self.use_weighted_mean_ap = use_weighted_mean_ap
self.label_id_offset = label_id_offset
self.groundtruth_boxes = {}
self.groundtruth_class_labels = {}
self.groundtruth_masks = {}
self.groundtruth_is_difficult_list = {}
self.groundtruth_is_group_of_list = {}
self.num_gt_instances_per_class = np.zeros(self.num_class, dtype=float)
self.num_gt_imgs_per_class = np.zeros(self.num_class, dtype=int)
self._initialize_detections()
def test_tp_fp(self):
num_groundtruth_classes = 3
matching_iou_threshold = 0.5
nms_iou_threshold = 1.0
nms_max_output_boxes = 10000
eval1 = per_image_evaluation.PerImageEvaluation(
num_groundtruth_classes, matching_iou_threshold, nms_iou_threshold,
nms_max_output_boxes)
detected_boxes = np.array(
[[0, 0, 1, 1], [10, 10, 5, 5], [0, 0, 2, 2], [5, 10, 10, 5],
[10, 5, 5, 10], [0, 0, 3, 3]],
dtype=float)
detected_scores = np.array([0.8, 0.1, 0.8, 0.9, 0.7, 0.8], dtype=float)
detected_class_labels = np.array([0, 1, 1, 2, 0, 2], dtype=int)
groundtruth_boxes = np.array([[0, 0, 1, 1], [0, 0, 3.5, 3.5]],
dtype=float)
groundtruth_class_labels = np.array([0, 2], dtype=int)
groundtruth_groundtruth_is_difficult_list = np.zeros(2, dtype=float)
groundtruth_groundtruth_is_group_of_list = np.array([False, False],
dtype=bool)
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,
groundtruth_groundtruth_is_group_of_list)
expected_scores = [np.array([0.8], dtype=float)] * 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 setUp(self):
num_groundtruth_classes = 1
matching_iou_threshold = 0.5
nms_iou_threshold = 1.0
nms_max_output_boxes = 10000
self.group_of_weight = 0.5
self.eval = per_image_evaluation.PerImageEvaluation(
num_groundtruth_classes, matching_iou_threshold, nms_iou_threshold,
nms_max_output_boxes, self.group_of_weight)
self.detected_boxes = np.array(
[[0, 0, 1, 1], [0, 0, 2, 1], [0, 0, 3, 1]], dtype=float)
self.detected_scores = np.array([0.8, 0.6, 0.5], dtype=float)
detected_masks_0 = np.array([[0, 1, 1, 0], [0, 0, 1, 0], [0, 0, 0, 0]],
dtype=np.uint8)
detected_masks_1 = np.array([[1, 0, 0, 0], [1, 1, 0, 0], [0, 0, 0, 0]],
dtype=np.uint8)
detected_masks_2 = np.array([[0, 0, 0, 0], [0, 1, 1, 0], [0, 1, 0, 0]],
dtype=np.uint8)
self.detected_masks = np.stack(
[detected_masks_0, detected_masks_1, detected_masks_2], axis=0)
self.groundtruth_boxes = np.array(
[[0, 0, 1, 1], [0, 0, 5, 5], [10, 10, 20, 20]], dtype=float)
groundtruth_masks_0 = np.array(
[[1, 0, 0, 0], [1, 0, 0, 0], [1, 0, 0, 0]], dtype=np.uint8)
groundtruth_masks_1 = np.array(
[[0, 0, 1, 0], [0, 0, 1, 0], [0, 0, 1, 0]], dtype=np.uint8)
groundtruth_masks_2 = np.array(
[[0, 1, 0, 0], [0, 1, 0, 0], [0, 1, 0, 0]], dtype=np.uint8)
self.groundtruth_masks = np.stack(
[groundtruth_masks_0, groundtruth_masks_1, groundtruth_masks_2],
axis=0)
def test_compute_corloc_with_very_large_iou_threshold(self):
num_groundtruth_classes = 3
matching_iou_threshold = 0.9
nms_iou_threshold = 1.0
nms_max_output_boxes = 10000
eval1 = per_image_evaluation.PerImageEvaluation(
num_groundtruth_classes, matching_iou_threshold, nms_iou_threshold,
nms_max_output_boxes)
detected_boxes = np.array(
[[0, 0, 1, 1], [0, 0, 2, 2], [0, 0, 3, 3], [0, 0, 5, 5]],
dtype=float)
detected_scores = np.array([0.9, 0.9, 0.1, 0.9], dtype=float)
detected_class_labels = np.array([0, 1, 0, 2], dtype=int)
groundtruth_boxes = np.array(
[[0, 0, 1, 1], [0, 0, 3, 3], [0, 0, 6, 6]], dtype=float)
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, 0], dtype=int)
self.assertTrue(
np.array_equal(expected_result,
is_class_correctly_detected_in_image))