Coverage improvements, MLE set to NotImplementedError (better in a new PR). (468d5bdf) · Commits · aflab / astrogeology / Autocnet

autocnet/camera/tests/test_utils.py

0 → 100644

+24 −0

Original line number	Diff line number	Diff line
		import os
		import sys
		import unittest

		import numpy as np

		sys.path.append(os.path.abspath('..'))

		from .. import utils


		class TestUtils(unittest.TestCase):

		def test_normalize(self):
		r = np.random.RandomState(12345)
		coords = r.uniform(0,100,size=(100,2))

		normalizer = utils.normalize(coords)

		truth = np.array([[0.036613, 0, -1.995647],
		[0, 0.036613, -1.862705],
		[0, 0, 1]])

		np.testing.assert_array_almost_equal(normalizer, truth)

autocnet/camera/utils.py

0 → 100644

+34 −0

Original line number	Diff line number	Diff line
		import math
		import numpy as np


		def normalize(a):
		"""
		Normalize a set of coordinates such that the origin is
		translated to the center and then scaled isotropically such
		that the average distance from the origin is $\sqrt{2}$.

		Parameters
		----------
		a : arraylike
		(n,2) array of x,y or (n,3) homogeneous coordinates

		Returns
		-------
		normalizer : ndarray
		(3,3) transformation matrix
		"""

		a = np.asarray(a)

		# Compute the normalization matrix
		centroid = a[:, :2].mean(axis=0)
		dist = np.sqrt(np.sum(((a[:, :2] - centroid)**2), axis=1))
		mean_dist = np.mean(dist)
		sq2 = math.sqrt(2)

		normalizer = np.array([[sq2 / mean_dist, 0, -sq2 / mean_dist * centroid[0]],
		[0, sq2 / mean_dist, -sq2 / mean_dist * centroid[1]],
		[0, 0, 1]])

		return normalizer
		No newline at end of file

autocnet/graph/node.py

+2 −7

Original line number	Diff line number	Diff line
		@@ -179,8 +179,6 @@ class Node(dict, MutableMapping):
		return self._keypoints.loc[index]
		else:
		return self._keypoints
		else:
		return None

		def get_keypoint_coordinates(self, index=None, homogeneous=False):
		"""
		@@ -425,11 +423,8 @@ class Node(dict, MutableMapping):
		if not hasattr(self, '_keypoints'):
		raise AttributeError('Keypoints must be extracted already, they have not been.')

		if clean_keys:
		mask = np.prod([self._mask_arrays[i] for i in clean_keys], axis=0, dtype=np.bool)
		keypoints = self._keypoints[mask]

		keypoints = self._keypoints[['x', 'y']].values
		matches, mask = self._clean(clean_keys)
		keypoints = self._keypoints[mask][['x', 'y']].values

		ratio = convex_hull_ratio(keypoints, ideal_area)
		return ratio

autocnet/transformation/transformations.py

+12 −33

Original line number	Diff line number	Diff line
		@@ -148,15 +148,15 @@ class TransformationMatrix(np.ndarray):

		@abc.abstractmethod
		def compute_error(self, x1, x2, index=None):
		pass
		raise NotImplementedError

		@abc.abstractmethod
		def recompute_matrix(self):
		pass
		raise NotImplementedError

		@abc.abstractmethod
		def refine(self):
		pass
		raise NotImplementedError


		class FundamentalMatrix(TransformationMatrix):
		@@ -196,7 +196,14 @@ class FundamentalMatrix(TransformationMatrix):
		----------
		.. [Hartley2003]
		"""

		raise NotImplementedError
		"""
		'''
		This still requires additional work.
		- The optimization is exceptionally slow.
		- Iteration is required to add newly discovered correspondences, re-estimate F using MLE,
		and continuing until the number of correspondences stabilizes.
		'''
		p = camera.idealized_camera()
		p1 = camera.estimated_camera_from_f(self)

		@@ -224,6 +231,7 @@ class FundamentalMatrix(TransformationMatrix):
		t = pgs[:, 3]
		M = pgs[:, 0:3]
		self[:] = crossform(t).dot(M)
		"""

		def refine(self, method=ps.esda.mapclassify.Fisher_Jenks, values=None, bin_id=0, **kwargs):
		"""
		@@ -341,35 +349,6 @@ class FundamentalMatrix(TransformationMatrix):

		return F_error

		def _normalize(self, a):
		"""
		Normalize a set of coordinates such that the origin is
		translated to the center and then scaled isotropically such
		that the average distance from the origin is $\sqrt{2}$.

		Parameters
		----------
		a : DataFrame
		(n,3) of homogeneous coordinates

		Returns
		-------
		normalized : ndarray
		(3,3) tranformation matrix
		"""

		# Compute the normalization matrix
		centroid = a[['x', 'y']].mean()
		dist = np.sqrt(np.sum(((a[['x', 'y']] - centroid)**2).values, axis=1))
		mean_dist = np.mean(dist)
		sq2 = math.sqrt(2)

		normalizer = np.array([[sq2 / mean_dist, 0, -sq2 / mean_dist * centroid[0]],
		[0, sq2 / mean_dist, -sq2 / mean_dist * centroid[1]],
		[0, 0, 1]])

		return normalizer

		def compute(self, kp1, kp2, method='ransac', reproj_threshold=2.0, confidence=0.99):
		"""
		Given two arrays of keypoints compute the fundamental matrix

functional_tests/test_three_image.py

+2 −2

Original line number	Diff line number	Diff line
		@@ -4,7 +4,6 @@ from autocnet.examples import get_path
		from autocnet.fileio.io_controlnetwork import to_isis
		from autocnet.fileio.io_controlnetwork import write_filelist
		from autocnet.graph.network import CandidateGraph
		from autocnet.matcher.matcher import FlannMatcher


		class TestThreeImageMatching(unittest.TestCase):
		@@ -54,6 +53,7 @@ class TestThreeImageMatching(unittest.TestCase):
		edge.ratio_check(clean_keys=['symmetry'], ratio=0.99)

		cg.apply_func_to_edges("compute_homography", clean_keys=['symmetry', 'ratio'])
		cg.compute_fundamental_matrices(clean_keys=['symmetry', 'ratio'])

		# Step: And create a C object
		cg.generate_cnet(clean_keys=['symmetry', 'ratio', 'ransac'])
		@@ -63,7 +63,7 @@ class TestThreeImageMatching(unittest.TestCase):
		write_filelist(filelist, 'TestThreeImageMatching_fromlist.lis')

		# Step: Create a correspondence network
		cg.generate_cnet(clean_keys=['symmetry', 'ratio', 'ransac'])
		cg.generate_cnet(clean_keys=['symmetry', 'ratio', 'ransac'], deepen=True)

		to_isis('TestThreeImageMatching.net', cg, mode='wb',
		networkid='TestThreeImageMatching', targetname='Moon')