def test_iou_should_return_1_when_masks_are_identical(self):
# Given
mask = np.random.randint(2, size=(2, 4))
# When
iou_result = iou(mask, mask)
# Then
expected_result = 1.0
self.assertEqual(expected_result, iou_result)
def test_iou_should_return_0_when_masks_are_opposites(self):
# Given
mask1 = np.random.randint(2, size=(2, 4))
mask2 = np.ones((2, 4)) - mask1
# When
iou_result = iou(mask1, mask2)
# Then
expected_result = 0.0
self.assertEqual(expected_result, iou_result)
def test_iou_should_return_0_25_when_mask_2_is_included_in_mask_1_and_masks_have_25_percent_overlap(self):
# Given
mask1 = np.ones((2, 4))
mask2 = np.concatenate((np.ones((2, 1)), np.zeros((2, 3))), axis=1)
# When
iou_result = iou(mask1, mask2)
# Then
expected_result = 0.25
self.assertEqual(expected_result, iou_result)
def test_iou_should_return_x_when_mask_2_is_not_included_in_mask_1_and_common_areas_are_50_percent_of_each_mask(
self):
# Given
mask1 = np.concatenate((np.zeros((2, 1)), np.ones((2, 2)), np.zeros((2, 1))), axis=1)
mask2 = np.concatenate((np.ones((2, 2)), np.zeros((2, 2))), axis=1)
# When
iou_result = iou(mask1, mask2)
# Then
expected_result = 1/3.0
self.assertEqual(expected_result, iou_result)
def compute_iou_for_query(batch_labels, batches, output, stick, testname):
classiou = 0
for i in range(0, batches.size()[0]):
# get prediction
pred = output.data.cpu().numpy()[i][0]
pred[pred <= 0.5] = 0
pred[pred > 0.5] = 1
# vis
demo = cv2.cvtColor(pred, cv2.COLOR_GRAY2RGB) * 255
stick[224 * 3:224 * 4, 224 * i:224 * (i + 1), :] = demo.copy()
testlabel = batch_labels.numpy()[i][0].astype(bool)
testlabel[testlabel <= 0.5] = 0
testlabel[testlabel > 0.5] = 1
# compute IOU
iou_score = iou(pred, testlabel)
classiou += iou_score
classiou /= 1.0 * batches.size()[0]
#print('iou=%0.4f for %s' % (classiou, testname))
return classiou, stick, i