Commit f3f02b07 authored by kberry's avatar kberry
Browse files

Updated outlier_detector to return a list which will mask the keypoints list... 

Updated outlier_detector to return a list which will mask the keypoints list and update tests accordingly
parent 74e53e3f

autocnet/matcher/outlier_detector.py

+27 −8
Original line number Diff line number Diff line
@@ -150,11 +150,27 @@ def compute_homography(kp1, kp2, outlier_algorithm=cv2.RANSAC, reproj_threshold= 
    mask = mask.astype(bool)

    return transformation_matrix, mask



# TODO: CITATION and better design?

def adaptive_non_max_suppression(keypoints, n=100, robust=0.9):

    """

    Select the top n keypoints, using Adaptive Non-Maximal Suppression

    to rank the keypoints in order of largest minimum suppression

    radius. A mask with only the positions of the top n keypoints set to 1 (and all else set to 0) is returned.



    Parameters

    ----------

    keypoints : list

               List of KeyPoint objects from a node of the graph or equivalently, for 1 image.



# TODO: document me

# right now, this is going to work on the un-merged dataframes... NOPE. I confused what was where again. Need the

# graph for keypoint information...

def adaptive_non_max_suppression(keypoints, n=20, robust=0.9):

    n : int

        The number of top-ranked keypoints to return.



    Returns

    -------

    keypoint_mask : list

                    A list containing a 1 in the positions of the top n selected keypoints and 0 in the positions

                    of all the other keypoints.

    """

    minimum_suppression_radius = {}

    for i, kp1 in enumerate(keypoints):

        x1, y1 = kp1.pt
@@ -166,10 +182,13 @@ def adaptive_non_max_suppression(keypoints, n=20, robust=0.9): 
        if(len(temp) > 0):

            minimum_suppression_radius[i] = np.min(np.array(temp))

        else:

            minimum_suppression_radius[i] = [np.nan]

    df = pd.DataFrame(list(minimum_suppression_radius.items()), columns=['keypoint_number', 'radius'])

    new_df = df.sort_values(by='radius', ascending=False).head(n)

    return new_df

            minimum_suppression_radius[i] = np.nan

    df = pd.DataFrame(list(minimum_suppression_radius.items()), columns=['keypoint', 'radius'])

    top_n = df.sort_values(by='radius', ascending=False).head(n)

    temp_df = df.mask(df.radius < top_n.radius.min(), other=np.nan)

    temp_df = temp_df.where(np.isnan(temp_df.keypoint), other=1)

    temp_df = temp_df.mask(np.isnan(temp_df.keypoint), other=0)

    return list(temp_df.radius)

functional_tests/test_two_image.py

+3 −3
Original line number Diff line number Diff line
@@ -49,10 +49,10 @@ class TestTwoImageMatching(unittest.TestCase): 
        for node, attributes in cg.nodes_iter(data=True):

            self.assertIn(len(attributes['keypoints']), range(490, 511))



        # Step: apply ANMS

        # Step: apply Adaptive non-maximal suppression

        for node, attributes in cg.nodes_iter(data=True):

            print(od.adaptive_non_max_suppression(attributes['keypoints']))

        self.assertTrue(False)

            attributes['keypoint_mask'] = od.adaptive_non_max_suppression(attributes['keypoints'])

            self.assertEqual(len(attributes['keypoint_mask']), len(attributes['keypoints']))



        # Step: Then apply a FLANN matcher

        fl = FlannMatcher()