def inference(image, score, output_file):
building_score = score[1]
building_mask_pred = (np.argmax(score, axis=0) == 1)
polygons = Mask(building_mask_pred).polygons()
new_predictions = []
for poly in polygons:
if len(poly) >= 3:
f = poly.reshape(-1, 2)
simplified_vw = simplify_coords_vwp(f, .3)
if len(simplified_vw) > 2:
mpoly = []
# Rebuilding the polygon in the way that PIL expects the values [(x1,y1),(x2,y2)]
for i in simplified_vw:
mpoly.append((i[0], i[1]))
# Adding the first point to the last to close the polygon
mpoly.append((simplified_vw[0][0], simplified_vw[0][1]))
new_predictions.append(mpoly)
# Creating the json with the predicted and then adjusted polygons
output_json = create_json(new_predictions)
with open(output_file, 'w') as out_file:
json.dump(output_json, out_file)
def inference2(image, weights, mean):
mean = np.load(mean)
model = segmentation_cpu.SegmentationModel(weights, mean)
image = np.array(Image.open(image))
score = model.apply_segmentation(image)
building_score = score[1]
building_mask_pred = (np.argmax(score, axis=0) == 1)
polygons = Mask(building_mask_pred).polygons()
new_predictions = []
for poly in polygons:
if len(poly) >= 3:
f = poly.reshape(-1, 2)
simplified_vw = simplify_coords_vwp(f, .3)
if len(simplified_vw) > 2:
mpoly = []
# Rebuilding the polygon in the way that PIL expects the values [(x1,y1),(x2,y2)]
for i in simplified_vw:
mpoly.append((i[0], i[1]))
# Adding the first point to the last to close the polygon
mpoly.append((simplified_vw[0][0], simplified_vw[0][1]))
new_predictions.append(mpoly)
# Creating the json with the predicted and then adjusted polygons
output_json = create_json(new_predictions)
return output_json
def predict_polygons(polygons):
new_predictions = []
for poly in polygons:
if len(poly) >= 3:
f = poly.reshape(-1, 2)
simplified_vw = simplify_coords_vwp(f, .3)
if len(simplified_vw) > 2:
mpoly = []
# Rebuilding the polygon in the way that PIL expects the values [(x1,y1),(x2,y2)]
for i in simplified_vw:
mpoly.append((i[0], i[1]))
# Adding the first point to the last to close the polygon
mpoly.append((simplified_vw[0][0], simplified_vw[0][1]))
new_predictions.append(mpoly)
return new_predictions
def mask_to_polygons(mask, use_convex_hull=False):
"""
convert predicted mask to polygons
"""
# for cv2 versions that do not support incontiguous array
mask = np.ascontiguousarray(mask)
mask = mask.astype(np.uint8)
# cv2.RETER_CCOMP flag retrieves all the contours
# the arranges them to a 2-level hierarchy.
res = cv2.findContours(mask, cv2.RETR_CCOMP, cv2.CHAIN_APPROX_SIMPLE)
hierarchy = res[-1]
# mask is empty
if hierarchy is None:
return [], False
has_holes = (hierarchy.reshape(-1, 4)[:, 3] >= 0).sum() > 0
res = res[-2]
try:
res_simp = simplify_coords_vwp(res[0].squeeze(), 30.0)
res_simp = np.array(res_simp)
res = [np.expand_dims(res_simp, axis=1)]
except ValueError:
print('Failed to simplify the points.')
if use_convex_hull:
hull = []
for i in range(len(res)):
hull.append(cv2.convexHull(res[i], False))
res = [x.flatten() for x in hull]
else:
res = [x.flatten() for x in res]
# convert OpenCV int coordinates [0, H -1 or W-1] to
# real value coordinate spaces
res = [x + 0.5 for x in res if len(x) >= 6]
return res, has_holes
def findContours(edges, min_arclength, contour_roughness):
contour_points, h = cv.findContours(edges, cv.RETR_TREE,
cv.CHAIN_APPROX_SIMPLE)
print("inital amount of cnts", len(contour_points))
#remove all contours with arclength less than 3 * edge length
cnts = []
for contour in contour_points:
if cv.arcLength(contour, True) > min_arclength * 3:
cnts.append(contour)
print("removed", len(contour_points) - len(cnts), "short contours")
contour_points = cnts
#simplify polygons using a topology preserving variant of the Visvalingam-Whyatt Algorithm
contour_points = [
simplify_coords_vwp(np.squeeze(contour), contour_roughness)
for contour in contour_points
]
# delete contours with less than 3 points (not even forming a polygon)
cnts = []
for contour in contour_points:
if len(contour) > 2:
cnts.append(contour)
print("removed",
len(contour_points) - len(cnts), " contours with less than 3 points")
contour_points = cnts
#turn contours into polygon objects
contour_points = [Polygon(contour) for contour in contour_points]
#delete invalid contours (non-closing)
cnts = []
invalid_contours = []
for polygon in contour_points:
if polygon.is_valid:
cnts.append(polygon)
else:
invalid_contours.append(polygon)
print("removed",
len(contour_points) - len(cnts), " contours that were invalid")
contour_points = cnts
#delete overlapping contours
cnts = []
overlap_cnts = []
for i in range(len(contour_points)):
a = contour_points[i]
if a not in overlap_cnts:
cnts.append(a)
for j in range(i, len(contour_points)):
b = contour_points[j]
if a != b:
if a.overlaps(b):
overlap_cnts.append(b)
print("removed",
len(contour_points) - len(cnts), " contours that were overlapping")
contour_points = [cnt for cnt in cnts if cnt not in overlap_cnts]
print("found", len(contour_points), "valid contours")
return contour_points, invalid_contours