def test_returns_number_of_intersections_for_different_segments(self):
result = max_segment_intersections(
[Segment(1, 4),
Segment(3, 5),
Segment(2, 7),
Segment(5, 10)])
self.assertEqual(result, 3)
def test_returns_for_points_on_different_values(self):
self.assertEqual([1, 0, 0], count_segments_for_points(
[Segment(0, 5), Segment(7, 10)], [1, 6, 11]))
self.assertEqual([0, 0, 1], count_segments_for_points(
[Segment(-10, 10)], [-100, 100, 0]))
self.assertEqual([2, 0], count_segments_for_points(
[Segment(0, 5), Segment(-3, 2), Segment(7, 10)], [1, 6]))
def extract_segment(self):
size = 10
step = 5
epsilon = 0.00089443
df1 = self.race1.df
df2 = self.race2.df
c1 = df1[[
"timestamp", "position_lat", "position_long", "altitude",
"distance"
]].values.T
c2 = df2[[
"timestamp", "position_lat", "position_long", "altitude",
"distance"
]].values.T
c1_matched_c2 = np.zeros(c1.shape[1])
mean_trace = np.zeros(
(4, c1.shape[1])) # [pos_lat, pos_long, altitude, distance]
timestamps_trace = np.zeros((2, c1.shape[1]), dtype='object')
for i2 in range(0, c2.shape[1] - size, step):
for i1 in range(0, c1.shape[1] - size, 1):
dist2 = np.linalg.norm(c1[1:3, i1:i1 + size] -
c2[1:3, i2:i2 + size])
if dist2 <= epsilon:
c1_matched_c2[i1:i1 + size] += 1
mean_trace[:3, i1:i1 + size] = (c1[1:4, i1:i1 + size] +
c2[1:4, i2:i2 + size]) / 2
# pour la distance on ne calcule pas la moyenne --> ça fausse tout
mean_trace[3, i1:i1 + size] = c1[4, i1:i1 + size]
timestamps_trace[:, i1:i1 + size] = [
c1[0, i1:i1 + size], c2[0, i2:i2 + size]
]
# Segment extractor (list of segment)
where = np.where(c1_matched_c2 > 1)[0]
diff = np.diff(where) > 1
if diff is False:
splitted = [where]
else:
sep = np.argwhere(diff).T[0]
splitted = np.split(where, sep + 1)
ret = [mean_trace[:, i] for i in splitted]
times = [timestamps_trace[:, i] for i in splitted]
# Drop segments shorter than 20
segments_filtered = [i for i in ret if i.shape[1] > 40]
times_filtered = [i for i in times if i.shape[1] > 40]
# drop les segments qui ne sont pas très correct (match mais pas de façon continue)
retour = []
times_2 = []
for segment, time in zip(segments_filtered, times_filtered):
diff = np.diff(segment[3, :])
if np.max(diff) < 50:
# valide si moins de 50m entre 2 points
retour.append(segment)
times_2.append(time)
# Pack for storage
for seg in zip(segments_filtered, times_filtered):
positions = seg[0]
timestamps1 = seg[1][0]
timestamps2 = seg[1][1]
self.segments.append(
Segment(positions, timestamps1, timestamps2, self.race1.points,
self.race2.points, "primitive"))
def load_segments(task_file, audio_folder, max_time):
segments = []
for segment_json in task_file["segments"]:
segments.append(Segment.from_json(segment_json, audio_folder, max_time))
return segments
def generate_breath_pause(self):
return Segment('silent', '', '', AudioSegment.silent(self.length), None, [], [])
def test_returns_empty_list_for_no_points(self):
self.assertEqual([], count_segments_for_points(
[Segment(0, 5), Segment(6, 7)], []))
def test_returns_for_points_on_border_of_segments(self):
self.assertEqual([1, 3, 3, 1, 2], count_segments_for_points(
[Segment(0, 5), Segment(2, 10), Segment(2, 7)], [0, 2, 5, 10, 7]))
def test_returns_for_points_on_the_segments(self):
self.assertEqual([1, 1, 1, 0], count_segments_for_points(
[Segment(0, 5), Segment(2, 7), Segment(10, 15)], [1, 6, 12, 8]))
def test_returns_for_single_point_on_the_intersection_of_segments(self):
self.assertEqual([2], count_segments_for_points(
[Segment(0, 5), Segment(2, 7)], [3]))
def test_returns_zero_for_single_point_on_the_right_of_segments(self):
self.assertEqual([0], count_segments_for_points(
[Segment(0, 5), Segment(2, 7)], [10]))
def test_returns_two_for_two_segments_with_intersections(self):
result = max_segment_intersections([Segment(1, 3), Segment(2, 5)])
self.assertEqual(result, 2)
def test_returns_one_for_single_segment(self):
result = max_segment_intersections([Segment(1, 3)])
self.assertEqual(result, 1)
def extract_segment(self, segment_type):
size = 10
step = 5
epsilon = 0.00089443
c1 = self.segment1.positions
c2 = self.segment2.positions
c1_matched_c2 = np.zeros(c1.shape[1])
mean_trace = np.zeros((4, c1.shape[1]))
timestamps_trace = np.zeros((2, c1.shape[1]), dtype='object')
for i2 in range(0, c2.shape[1] - size, step):
for i1 in range(0, c1.shape[1] - size, 1):
dist2 = np.linalg.norm(c1[0:2, i1:i1 + size] -
c2[0:2, i2:i2 + size])
if dist2 <= epsilon:
c1_matched_c2[i1:i1 + size] += 1
mean_trace[:3, i1:i1 + size] = (c1[0:3, i1:i1 + size] +
c2[0:3, i2:i2 + size]) / 2
mean_trace[3, i1:i1 + size] = c1[3, i1:i1 + size]
timestamps_trace[:, i1:i1 + size] = [
self.segment1.times1[i1:i1 + size],
self.segment2.times1[i2:i2 + size]
]
# Segment extractor (list of segment)
where = np.where(c1_matched_c2 > 1)[0]
diff = np.diff(where) > 1
if diff is False:
splitted = [where]
else:
sep = np.argwhere(diff).T[0]
splitted = np.split(where, sep + 1)
ret = [mean_trace[:, i] for i in splitted]
times = [timestamps_trace[:, i] for i in splitted]
# Drop segments shorter than 20
segments_filtered = [i for i in ret if i.shape[1] > 40]
times_filtered = [i for i in times if i.shape[1] > 40]
retour = []
times_2 = []
for segment, time in zip(segments_filtered, times_filtered):
diff = np.diff(segment[3, :])
if np.max(diff) < 50:
retour.append(segment)
times_2.append(time)
# Pack for storage
for seg in zip(segments_filtered, times_filtered):
positions = seg[0]
timestamps1 = seg[1][0]
timestamps2 = seg[1][1]
self.segments.append(
Segment(positions, timestamps1, timestamps2,
self.segment1.points1, self.segment2.points1,
segment_type))