def query(self, picture: Picture) -> List[Picture]:
mask, rec = detect_text(picture.get_image())
kp, des = self.sift.detectAndCompute(picture.get_image(), mask)
return (seq(
self.db).map(lambda p: (p[0], self.bf.knnMatch(p[2], des, k=2))).
map(lambda p: (p[0], self._ratio_test(p[1]))).map(lambda p: (p[
0], len(p[1]))).filter(lambda p: p[1] > 0).sorted(
lambda p: p[1],
reverse=True).map(lambda p: p[0]).take(10).to_list())
def query(self, picture: Picture) -> List[Picture]:
mask, rec = detect_text(picture.get_image())
kp, des = self.surf.detectAndCompute(picture.get_image(), mask)
return (
seq(self.db).map(lambda p: (p[0], self.bf.match(p[2], des))).
map(lambda p: (p[0], seq(p[1]).filter(lambda d: d.distance < max(
THRESHOLD,
seq(p[1]).map(lambda m: m.distance).min())).to_list())).
map(lambda p: (p[0], len(p[1]))).filter(lambda p: p[1] > 4).sorted(
lambda p: p[1],
reverse=True).map(lambda p: p[0]).take(10).to_list())
def query(self, picture: Picture) -> List[Picture]:
mask, rec = detect_text(picture.get_image())
kp, des = self.sift.detectAndCompute(picture.get_image(), mask)
FLANN_INDEX_KDTREE = 1
flann_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)
flann = cv2.FlannBasedMatcher(flann_params, {})
return (seq(self.db).map(
lambda p: (p[0],
self.flann.knnMatch(np.asarray(p[2], np.float32),
np.asarray(des, np.float32), 2))
).map(lambda p: (p[0], self._ratio_test(p[1]))).map(
lambda p: (p[0], len(p[1]))).filter(lambda p: p[1] > 4).sorted(
lambda p: p[1],
reverse=True).map(lambda p: p[0]).take(10).to_list())
def query(self, picture: Picture) -> List[Picture]:
mask,rec=detect_text(picture.get_image())
kp, des = self.orb.detectAndCompute(picture.get_image(), mask)
return (
seq(self.db)
.map(lambda p: (p[0], p[1], self.bf.knnMatch(des, p[2], k=2)))
.map(lambda p: (p[0], p[1], self._ratio_test(p[2])))
.filter(lambda p: len(p[2]) > MIN_MATCH_COUNT)
.map(lambda p: (p[0], self._homography(kp, p[1], p[2])))
.filter(lambda p: len(p[1]) > MIN_MATCH_COUNT)
.sorted(lambda p: p[1], reverse=True)
.map(lambda p: p[0])
.take(10)
.to_list()
)
def query(self, picture: Picture) -> List[Picture]:
mask, rec = detect_text(picture.get_image())
kp, des = self.orb.detectAndCompute(picture.get_image(), mask)
return (seq(
self.db).map(lambda p: (p[0], p[1], self.bf.match(des, p[2]))).map(
lambda p:
(p[0], p[1], seq(p[2]).filter(lambda d: d.distance < max(
THRESHOLD,
seq(p[2]).map(lambda m: m.distance).min())).to_list())).
filter(lambda p: len(p[2]) > MIN_MATCH_COUNT).map(
lambda p: (p[0],
self._homography(kp, p[1], p[2], p[0].get_image(
), picture.get_image()))).
filter(lambda p: (p[1]) > MIN_MATCH_COUNT).sorted(
lambda p: p[1],
reverse=True).map(lambda p: p[0]).take(10).to_list())
def query(self, picture: Picture) -> List[Picture]:
mask, rec = detect_text(picture.get_image())
kp, des = self.orb.detectAndCompute(picture.get_image(), mask)
return (
seq(self.db)
.map(
lambda p: (p[0], self.flann.knnMatch(np.asarray(p[2], np.float32), np.asarray(des, np.float32), 2)))
.map(lambda p: (p[0], self._ratio_test(p[1])))
.map(lambda p: (p[0], len(p[1])))
.filter(lambda p: p[1] > 4)
.sorted(lambda p: p[1], reverse=True)
.map(lambda p: p[0])
.take(10)
.to_list()
)
def query(self, picture: Picture) -> List[Picture]:
hist = self._get_histogram(picture.get_image())
return (seq(self.db)
# Calculate histogram similarity
.map(lambda entry:
(entry[0], self._compare_histograms_full(hist, entry[1])))
# Calculate distance to center
.map(lambda entry:
(entry[0], self._euclidean_distance_to_origin(entry[1])))
# Order by distance
.sorted(
lambda entry_res: entry_res[1], False if
self.histogram_comparison_method == cv2.HISTCMP_HELLINGER
else True).map(lambda entry: entry[0])
# Take first K
.take(K).to_list())
def query(self, picture: Picture, frame: Frame = None) -> List[Picture]:
im = picture.get_image()
if frame and frame.is_valid():
side = int(math.sqrt(frame.get_area()) * 0.8)
m = frame.get_perspective_matrix(
np.array([[0, side - 1], [side - 1, side - 1], [side - 1, 0],
[0, 0]]))
im = cv2.warpPerspective(im, m, (side, side))
# plt.imshow(cv2.cvtColor(im, cv2.COLOR_BGR2RGB))
# plt.show()
kp, des = self.orb.detectAndCompute(im, None)
return (
seq(self.db).map(lambda p: (p[0], self.bf.match(p[2], des))).
map(lambda p: (p[0], seq(p[1]).filter(lambda d: d.distance < max(
THRESHOLD,
seq(p[1]).map(lambda m: m.distance).min())).to_list())).
map(lambda p: (p[0], len(p[1]))).filter(lambda p: p[1] > 4).sorted(
lambda p: p[1],
reverse=True).map(lambda p: p[0]).take(10).to_list())
def query(self, picture: Picture) -> (List[Picture], Frame):
frame = get_frame_with_lines(picture.get_image())
return self.orb.query(picture), frame
