def paintCorners(painter, corners, rect):
# FIXME: This only works right for orthogonal maps right now
hx = rect.width() / 2
hy = rect.height() / 2
x = corners
if x==Corners.TopLeft:
painter.drawPie(rect.translated(-hx, -hy), -90 * 16, 90 * 16)
elif x==Corners.TopRight:
painter.drawPie(rect.translated(hx, -hy), 180 * 16, 90 * 16)
elif x==Corners.TopRight | Corners.TopLeft:
painter.drawRect(rect.x(), rect.y(), rect.width(), hy)
elif x==Corners.BottomLeft:
painter.drawPie(rect.translated(-hx, hy), 0, 90 * 16)
elif x==Corners.BottomLeft | Corners.TopLeft:
painter.drawRect(rect.x(), rect.y(), hx, rect.height())
elif x==Corners.BottomLeft | Corners.TopRight:
painter.drawPie(rect.translated(-hx, hy), 0, 90 * 16)
painter.drawPie(rect.translated(hx, -hy), 180 * 16, 90 * 16)
elif x==Corners.BottomLeft | Corners.TopRight | Corners.TopLeft:
fill = QPainterPath()
ellipse = QPainterPath()
fill.addRect(rect)
ellipse.addEllipse(rect.translated(hx, hy))
painter.drawPath(fill.subtracted(ellipse))
elif x==Corners.BottomRight:
painter.drawPie(rect.translated(hx, hy), 90 * 16, 90 * 16)
elif x==Corners.BottomRight | Corners.TopLeft:
painter.drawPie(rect.translated(-hx, -hy), -90 * 16, 90 * 16)
painter.drawPie(rect.translated(hx, hy), 90 * 16, 90 * 16)
elif x==Corners.BottomRight | Corners.TopRight:
painter.drawRect(rect.x() + hx, rect.y(), hx, rect.height())
elif x==Corners.BottomRight | Corners.TopRight | Corners.TopLeft:
fill = QPainterPath()
ellipse = QPainterPath()
fill.addRect(rect)
ellipse.addEllipse(rect.translated(-hx, hy))
painter.drawPath(fill.subtracted(ellipse))
elif x==Corners.BottomRight | Corners.BottomLeft:
painter.drawRect(rect.x(), rect.y() + hy, rect.width(), hy)
elif x==Corners.BottomRight | Corners.BottomLeft | Corners.TopLeft:
fill = QPainterPath()
ellipse = QPainterPath()
fill.addRect(rect)
ellipse.addEllipse(rect.translated(hx, -hy))
painter.drawPath(fill.subtracted(ellipse))
elif x==Corners.BottomRight | Corners.BottomLeft | Corners.TopRight:
fill = QPainterPath()
ellipse = QPainterPath()
fill.addRect(rect)
ellipse.addEllipse(rect.translated(-hx, -hy))
painter.drawPath(fill.subtracted(ellipse))
elif x==Corners.BottomRight | Corners.BottomLeft | Corners.TopRight | Corners.TopLeft:
painter.drawRect(rect)
def paintCorners(painter, corners, rect):
# FIXME: This only works right for orthogonal maps right now
hx = rect.width() / 2
hy = rect.height() / 2
x = corners
if x == Corners.TopLeft:
painter.drawPie(rect.translated(-hx, -hy), -90 * 16, 90 * 16)
elif x == Corners.TopRight:
painter.drawPie(rect.translated(hx, -hy), 180 * 16, 90 * 16)
elif x == Corners.TopRight | Corners.TopLeft:
painter.drawRect(rect.x(), rect.y(), rect.width(), hy)
elif x == Corners.BottomLeft:
painter.drawPie(rect.translated(-hx, hy), 0, 90 * 16)
elif x == Corners.BottomLeft | Corners.TopLeft:
painter.drawRect(rect.x(), rect.y(), hx, rect.height())
elif x == Corners.BottomLeft | Corners.TopRight:
painter.drawPie(rect.translated(-hx, hy), 0, 90 * 16)
painter.drawPie(rect.translated(hx, -hy), 180 * 16, 90 * 16)
elif x == Corners.BottomLeft | Corners.TopRight | Corners.TopLeft:
fill = QPainterPath()
ellipse = QPainterPath()
fill.addRect(rect)
ellipse.addEllipse(rect.translated(hx, hy))
painter.drawPath(fill.subtracted(ellipse))
elif x == Corners.BottomRight:
painter.drawPie(rect.translated(hx, hy), 90 * 16, 90 * 16)
elif x == Corners.BottomRight | Corners.TopLeft:
painter.drawPie(rect.translated(-hx, -hy), -90 * 16, 90 * 16)
painter.drawPie(rect.translated(hx, hy), 90 * 16, 90 * 16)
elif x == Corners.BottomRight | Corners.TopRight:
painter.drawRect(rect.x() + hx, rect.y(), hx, rect.height())
elif x == Corners.BottomRight | Corners.TopRight | Corners.TopLeft:
fill = QPainterPath()
ellipse = QPainterPath()
fill.addRect(rect)
ellipse.addEllipse(rect.translated(-hx, hy))
painter.drawPath(fill.subtracted(ellipse))
elif x == Corners.BottomRight | Corners.BottomLeft:
painter.drawRect(rect.x(), rect.y() + hy, rect.width(), hy)
elif x == Corners.BottomRight | Corners.BottomLeft | Corners.TopLeft:
fill = QPainterPath()
ellipse = QPainterPath()
fill.addRect(rect)
ellipse.addEllipse(rect.translated(hx, -hy))
painter.drawPath(fill.subtracted(ellipse))
elif x == Corners.BottomRight | Corners.BottomLeft | Corners.TopRight:
fill = QPainterPath()
ellipse = QPainterPath()
fill.addRect(rect)
ellipse.addEllipse(rect.translated(-hx, -hy))
painter.drawPath(fill.subtracted(ellipse))
elif x == Corners.BottomRight | Corners.BottomLeft | Corners.TopRight | Corners.TopLeft:
painter.drawRect(rect)
def paintEvent(self, event: QPaintEvent):
"""
重写绘图事件
:param event:
:return:
"""
try:
backPath = QPainterPath()
backPath.addRect(0, 0, self.width(), self.height())
fillPath = QPainterPath()
if hasattr(self, "startPoint") and hasattr(self, "endPoint"):
movePath = QPainterPath()
movePath.addRect(QRectF(self.startPoint, self.endPoint))
fillPath = backPath.subtracted(movePath)
else:
fillPath = backPath
# 创建绘图设备
painter = QPainter(self)
# 绘制背景图
painter.drawImage(QPoint(0, 0), self.backImg)
painter.setPen(QPen(QColor(87, 170, 255), 5, Qt.SolidLine))
# painter.drawPath(fillPath)
# 填充非选择区域
painter.fillPath(fillPath, QColor(0, 0, 0, 100))
except Exception as e:
traceback.print_exc()
def _dbgRingShape(self, p):
r, R, = self.boundings()
opath = QPainterPath()
opath.setFillRule(Qt.WindingFill)
opath.addEllipse(-R, -R, 2*R, 2*R)
ipath = QPainterPath()
ipath.setFillRule(Qt.WindingFill)
ipath.addEllipse(-r, -r, 2*r, 2*r)
p.fillPath(opath.subtracted(ipath), QColor(255, 255, 0, 50))
p.strokePath(opath.simplified(), QPen(Qt.black, 3))
p.strokePath(ipath, QPen(Qt.black, 1))
class Blob(object):
"""
Blob data. A blob is a group of pixels.
It can be tagged with the class and other information.
It is stored as an outer contour (the border) plus a list of inner contours (holes).
"""
def __init__(self, region, offset_x, offset_y, id):
self.version = 0
self.id = int(id)
self.area = 0.0
self.surface_area = 0.0
self.perimeter = 0.0
self.centroid = np.zeros((2))
self.bbox = np.zeros((4))
# placeholder; empty contour
self.contour = np.zeros((2, 2))
self.inner_contours = []
self.qpath = None
self.qpath_gitem = None
self.instance_name = "noname"
self.blob_name = "noname"
if region:
# extract properties
self.centroid = np.array(region.centroid)
cy = self.centroid[0]
cx = self.centroid[1]
self.centroid[0] = cx + offset_x
self.centroid[1] = cy + offset_y
# Bounding box (min_row, min_col, max_row, max_col).
# Pixels belonging to the bounding box are in the half-open
# interval [min_row, max_row) and [min_col, max_col).
self.bbox = np.array(region.bbox)
width = self.bbox[3] - self.bbox[1]
height = self.bbox[2] - self.bbox[0]
# BBOX -> TOP, LEFT, WIDTH, HEIGHT
self.bbox[0] = self.bbox[0] + offset_y
self.bbox[1] = self.bbox[1] + offset_x
self.bbox[2] = width
self.bbox[3] = height
# QPainterPath associated with the contours
self.qpath = None
# QGraphicsItem associated with the QPainterPath
self.qpath_gitem = None
# to extract the contour we use the mask cropped according to the bbox
input_mask = region.image.astype(int)
self.contour = np.zeros((2, 2))
self.inner_contours = []
self.updateUsingMask(self.bbox, input_mask)
# a string with a progressive number to identify the instance
self.instance_name = "coral" + str(id)
# a string with a number to identify the blob plus its centroid
xc = self.centroid[0]
yc = self.centroid[1]
self.blob_name = "c-{:d}-{:.1f}x-{:.1f}y".format(self.id, xc, yc)
# deep extreme points (for fine-tuning)
self.deep_extreme_points = np.zeros((4, 2))
# name of the class
self.class_name = "Empty"
# color of the class
self.class_color = [128, 128, 128]
self.genet = None
# note about the coral, i.e. damage type
self.note = ""
# QImage corresponding to the current mask
self.qimg_mask = None
# QPixmap associated with the mask (for pixel-level editing operations)
self.pxmap_mask = None
# QGraphicsItem associated with the pixmap mask
self.pxmap_mask_gitem = None
# membership group (if any)
self.group = None
def copy(self):
blob = Blob(None, 0, 0, 0)
blob.area = self.area
blob.surface_area = self.surface_area
blob.perimeter = self.perimeter
blob.centroid = self.centroid
blob.bbox = self.bbox
blob.contour = self.contour.copy()
for inner in self.inner_contours:
blob.inner_contours.append(inner.copy())
blob.qpath_gitem = None
blob.qpath = None
blob.instance_name = blob.instance_name
blob.blob_name = self.blob_name
blob.id = self.id
blob.version = self.version + 1
blob.genet = self.genet
blob.class_name = self.class_name
blob.deep_extreme_points = self.deep_extreme_points
self.note = ""
self.qimg_mask = None
self.pxmap_mask = None
self.pxmap_mask_gitem = None
return blob
def __deepcopy__(self, memo):
#avoid recursion!
deepcopy_method = self.__deepcopy__
self.__deepcopy__ = None
#save and later restore qobjects
path = self.qpath
pathitem = self.qpath_gitem
#no deep copy for qobjects
self.qpath = None
self.qpath_gitem = None
blob = copy.deepcopy(self)
blob.contour = self.contour.copy()
blob.inner_contours.clear()
for inner in self.inner_contours:
blob.inner_contours.append(inner.copy())
blob.qpath = None
blob.qpath_gitem = None
self.qpath = path
self.qpath_gitem = pathitem
#restore deepcopy (also to the newly created Blob!
blob.__deepcopy__ = self.__deepcopy__ = deepcopy_method
return blob
def setId(self, id):
# a string with a number to identify the blob plus its centroid
xc = self.centroid[0]
yc = self.centroid[1]
self.id = id
self.blob_name = "c-{:d}-{:.1f}x-{:.1f}y".format(self.id, xc, yc)
def getMask(self):
"""
It creates the mask from the contour and returns it.
"""
r = self.bbox[3]
c = self.bbox[2]
origin = np.array([int(self.bbox[1]), int(self.bbox[0])])
mask = np.zeros((r, c), np.uint8)
points = self.contour.round().astype(int)
fillPoly(mask, pts=[points - origin], color=(1))
# holes
for inner_contour in self.inner_contours:
points = inner_contour.round().astype(int)
fillPoly(mask, pts=[points - origin], color=(0, 0, 0))
return mask
def updateUsingMask(self, bbox, mask):
self.createContourFromMask(mask, bbox)
self.calculatePerimeter()
self.calculateCentroid(mask, bbox)
self.calculateArea(mask)
self.bbox = Mask.pointsBox(self.contour, 4)
def createFromClosedCurve(self, lines):
"""
It creates a blob starting from a closed curve. If the curve is not closed False is returned.
If the curve intersect itself many times the first segmented region is created.
"""
points = self.lineToPoints(lines)
box = Mask.pointsBox(points, 4)
(mask, box) = Mask.jointMask(box, box)
Mask.paintPoints(mask, box, points, 1)
before = np.count_nonzero(mask)
mask = ndi.binary_fill_holes(mask)
after = np.count_nonzero(mask)
if before == after:
return False
selem = np.array([[0, 0, 0], [0, 0, 1], [0, 1, 0]])
mask = binary_erosion(mask, selem)
self.updateUsingMask(box, mask)
return True
def createContourFromMask(self, mask, bbox):
"""
It creates the contour (and the corrisponding polygon) from the blob mask.
"""
# NOTE: The mask is expected to be cropped around its bbox (!!) (see the __init__)
self.inner_contours.clear()
# we need to pad the mask to avoid to break the contour that touches the borders
PADDED_SIZE = 4
img_padded = pad(mask, (PADDED_SIZE, PADDED_SIZE),
mode="constant",
constant_values=(0, 0))
contours = measure.find_contours(img_padded, 0.6)
inner_contours = measure.find_contours(img_padded, 0.4)
number_of_contours = len(contours)
threshold = 20 #min number of points in a small hole
if number_of_contours > 1:
# search the contour with the largest bounding box (area)
max_area = 0
longest = 0
for i, contour in enumerate(contours):
cbox = Mask.pointsBox(contour, 0)
area = cbox[2] * cbox[3]
if area > max_area:
max_area = area
longest = i
max_area = 0
inner_longest = 0
for i, contour in enumerate(inner_contours):
cbox = Mask.pointsBox(contour, 0)
area = cbox[2] * cbox[3]
if area > max_area:
max_area = area
inner_longest = i
# divide the contours in OUTER contour and INNER contours
for i, contour in enumerate(contours):
if i == longest:
self.contour = np.array(contour)
for i, contour in enumerate(inner_contours):
if i != inner_longest:
if contour.shape[0] > threshold:
coordinates = np.array(contour)
self.inner_contours.append(coordinates)
# adjust the coordinates of the outer contour
# (NOTE THAT THE COORDINATES OF THE BBOX ARE IN THE GLOBAL MAP COORDINATES SYSTEM)
for i in range(self.contour.shape[0]):
ycoor = self.contour[i, 0]
xcoor = self.contour[i, 1]
self.contour[i, 0] = xcoor - PADDED_SIZE + bbox[1]
self.contour[i, 1] = ycoor - PADDED_SIZE + bbox[0]
# adjust coordinates of the INNER contours
for j, contour in enumerate(self.inner_contours):
for i in range(contour.shape[0]):
ycoor = contour[i, 0]
xcoor = contour[i, 1]
self.inner_contours[j][i,
0] = xcoor - PADDED_SIZE + bbox[1]
self.inner_contours[j][i,
1] = ycoor - PADDED_SIZE + bbox[0]
elif number_of_contours == 1:
coords = measure.approximate_polygon(contours[0], tolerance=0.2)
self.contour = np.array(coords)
# adjust the coordinates of the outer contour
# (NOTE THAT THE COORDINATES OF THE BBOX ARE IN THE GLOBAL MAP COORDINATES SYSTEM)
for i in range(self.contour.shape[0]):
ycoor = self.contour[i, 0]
xcoor = self.contour[i, 1]
self.contour[i, 0] = xcoor - PADDED_SIZE + bbox[1]
self.contour[i, 1] = ycoor - PADDED_SIZE + bbox[0]
else:
raise Exception("Empty contour")
#TODO optimize the bbox
self.bbox = bbox
def lineToPoints(self, lines, snap=False):
points = np.empty(shape=(0, 2), dtype=int)
for line in lines:
p = self.drawLine(line)
if p.shape[0] == 0:
continue
if snap:
p = self.snapToBorder(p)
if p is None:
continue
points = np.append(points, p, axis=0)
return points
def dilate(self, size=1):
"""Dilate the blob"""
mask = self.getMask()
dilated_mask = binary_dilation(mask, square(size))
self.updateUsingMask(self.bbox, dilated_mask)
def erode(self, size=1):
"""Erode the blob"""
mask = self.getMask()
eroded_mask = binary_erosion(mask, square(size))
self.updateUsingMask(self.bbox, eroded_mask)
def drawLine(self, line):
(x, y) = utils.draw_open_polygon(line[:, 1], line[:, 0])
points = np.asarray([x, y]).astype(int)
points = points.transpose()
points[:, [1, 0]] = points[:, [0, 1]]
return points
def snapToBorder(self, points):
return self.snapToContour(points, self.contour)
def snapToContour(self, points, contour):
"""
Given a curve specified as a set of points, snap the curve on the blob mask:
1) the initial segments of the curve are removed until they snap
2) the end segments of the curve are removed until they snap
"""
test = points_in_poly(points, contour)
if test is None or test.shape[0] <= 3:
return None
jump = np.gradient(test.astype(int))
ind = np.nonzero(jump)
ind = np.asarray(ind)
snappoints = None
if ind.shape[1] > 2:
first_el = ind[0, 0]
last_el = ind[0, -1]
snappoints = points[first_el:last_el + 1, :].copy()
return snappoints
def snapToInternalBorders(self, points):
if not self.inner_contours:
return None
snappoints = np.zeros(shape=(0, 2))
for contour in self.inner_contours:
snappoints = np.append(snappoints,
self.snapToContour(points, contour))
return snappoints
def setupForDrawing(self):
"""
Create the QPolygon and the QPainterPath according to the blob's contours.
"""
# QPolygon to draw the blob
#working with mask the center of the pixels is in 0, 0
#if drawing the center of the pixel is 0.5, 0.5
qpolygon = QPolygonF()
for i in range(self.contour.shape[0]):
qpolygon << QPointF(self.contour[i, 0] + 0.5,
self.contour[i, 1] + 0.5)
self.qpath = QPainterPath()
self.qpath.addPolygon(qpolygon)
for inner_contour in self.inner_contours:
qpoly_inner = QPolygonF()
for i in range(inner_contour.shape[0]):
qpoly_inner << QPointF(inner_contour[i, 0], inner_contour[i,
1])
path_inner = QPainterPath()
path_inner.addPolygon(qpoly_inner)
self.qpath = self.qpath.subtracted(path_inner)
def createQPixmapFromMask(self):
w = self.bbox[2]
h = self.bbox[3]
self.qimg_mask = QImage(w, h, QImage.Format_ARGB32)
self.qimg_mask.fill(qRgba(0, 0, 0, 0))
if self.class_name == "Empty":
rgba = qRgba(255, 255, 255, 255)
else:
rgba = qRgba(self.class_color[0], self.class_color[1],
self.class_color[2], 100)
blob_mask = self.getMask()
for x in range(w):
for y in range(h):
if blob_mask[y, x] == 1:
self.qimg_mask.setPixel(x, y, rgba)
self.pxmap_mask = QPixmap.fromImage(self.qimg_mask)
#bbox is used to place the mask!
def calculateCentroid(self, mask, bbox):
m = measure.moments(mask)
c = np.array((m[0, 1] / m[0, 0], m[1, 0] / m[0, 0]))
#centroid is (x, y) while measure returns (y,x and bbox is yx)
self.centroid = np.array((c[0] + bbox[1], c[1] + bbox[0]))
self.blob_name = "c-{:d}-{:.1f}x-{:.1f}y".format(
self.id, self.centroid[0], self.centroid[1])
def calculateContourPerimeter(self, contour):
#self.perimeter = measure.perimeter(mask) instead?
# perimeter of the outer contour
px1 = contour[0, 0]
py1 = contour[0, 1]
N = contour.shape[0]
pxlast = contour[N - 1, 0]
pylast = contour[N - 1, 1]
perim = math.sqrt((px1 - pxlast) * (px1 - pxlast) + (py1 - pylast) *
(py1 - pylast))
for i in range(1, contour.shape[0]):
px2 = contour[i, 0]
py2 = contour[i, 1]
d = math.sqrt((px1 - px2) * (px1 - px2) + (py1 - py2) *
(py1 - py2))
perim += d
px1 = px2
py1 = py2
return perim
def calculatePerimeter(self):
#tole = 2
#simplified = measure.approximate_polygon(self.contour, tolerance=tole)
self.perimeter = self.calculateContourPerimeter(self.contour)
for contour in self.inner_contours:
#simplified = measure.approximate_polygon(contour, tolerance=tole)
self.perimeter += self.calculateContourPerimeter(contour)
def calculateArea(self, mask):
self.area = mask.sum().astype(float)
def fromDict(self, dict):
"""
Set the blob information given it represented as a dictionary.
"""
self.bbox = np.asarray(dict["bbox"])
self.centroid = np.asarray(dict["centroid"])
self.area = dict["area"]
self.perimeter = dict["perimeter"]
self.contour = np.asarray(dict["contour"])
inner_contours = dict["inner contours"]
self.inner_contours = []
for c in inner_contours:
self.inner_contours.append(np.asarray(c))
self.deep_extreme_points = np.asarray(dict["deep_extreme_points"])
self.class_name = dict["class name"]
self.class_color = dict["class color"]
self.instance_name = dict["instance name"]
self.blob_name = dict["blob name"]
self.id = int(dict["id"])
self.note = dict["note"]
def save(self):
return self.toDict()
def toDict(self):
"""
Get the blob information as a dictionary.
"""
dict = {}
dict["bbox"] = self.bbox.tolist()
dict["centroid"] = self.centroid.tolist()
dict["area"] = self.area
dict["perimeter"] = self.perimeter
dict["contour"] = self.contour.tolist()
dict["inner contours"] = []
for c in self.inner_contours:
dict["inner contours"].append(c.tolist())
dict["deep_extreme_points"] = self.deep_extreme_points.tolist()
dict["class name"] = self.class_name
dict["class color"] = self.class_color
dict["instance name"] = self.instance_name
dict["blob name"] = self.blob_name
dict["id"] = self.id
dict["note"] = self.note
return dict
class Blob(object):
"""
Blob data. A blob is a group of pixels.
A blob can be tagged with the class and other information.
Both the set of pixels and the corresponding vectorized version are stored.
"""
def __init__(self, region, offset_x, offset_y, id):
if region == None:
# AN EMPTY BLOB IS CREATED..
self.area = 0.0
self.perimeter = 0.0
self.centroid = np.zeros((2))
self.bbox = np.zeros((4))
# placeholder; empty contour
self.contour = np.zeros((2, 2))
self.inner_contours = []
self.qpath = None
self.qpath_gitem = None
self.instance_name = "noname"
self.blob_name = "noname"
self.id = 0
else:
# extract properties
self.centroid = np.array(region.centroid)
cy = self.centroid[0]
cx = self.centroid[1]
self.centroid[0] = cx + offset_x
self.centroid[1] = cy + offset_y
# Bounding box (min_row, min_col, max_row, max_col).
# Pixels belonging to the bounding box are in the half-open
# interval [min_row, max_row) and [min_col, max_col).
self.bbox = np.array(region.bbox)
width = self.bbox[3] - self.bbox[1]
height = self.bbox[2] - self.bbox[0]
# BBOX -> TOP, LEFT, WIDTH, HEIGHT
self.bbox[0] = self.bbox[0] + offset_y
self.bbox[1] = self.bbox[1] + offset_x
self.bbox[2] = width
self.bbox[3] = height
# QPainterPath associated with the contours
self.qpath = None
# QGraphicsItem associated with the QPainterPath
self.qpath_gitem = None
# to extract the contour we use the mask cropped according to the bbox
input_mask = region.image.astype(int)
self.contour = np.zeros((2, 2))
self.inner_contours = []
self.createContourFromMask(input_mask)
self.setupForDrawing()
self.calculatePerimeter()
self.calculateArea(input_mask)
# a string with a progressive number to identify the instance
self.instance_name = "coral" + str(id)
# a string with a number to identify the blob plus its centroid
xc = int(self.centroid[0])
yc = int(self.centroid[1])
self.blob_name = "blob" + str(id) + "-" + str(xc) + "-" + str(yc)
self.id = id
# deep extreme points (for fine-tuning)
self.deep_extreme_points = np.zeros((4, 2))
# name of the class
self.class_name = "Empty"
# color of the class
self.class_color = [128, 128, 128]
# note about the coral, i.e. damage type
self.note = ""
# QImage corresponding to the current mask
self.qimg_mask = None
# QPixmap associated with the mask (for pixel-level editing operations)
self.pxmap_mask = None
# QGraphicsItem associated with the pixmap mask
self.pxmap_mask_gitem = None
# membership group (if any)
self.group = None
def copy(self):
blob = Blob()
blob.centroid = self.centroid.copy()
blob.bbox = self.bbox.copy()
blob.classname = self.classname
blob.classcolor = self.classcolor
blob.instance_name = self.instance_name
blob.id = self.id
blob.note = self.note
return blob
def setId(self, id):
# a string with a number to identify the blob plus its centroid
xc = int(self.centroid[0])
yc = int(self.centroid[1])
self.blob_name = "blob" + str(id) + "-" + str(xc) + "-" + str(yc)
self.id = id
def getMask(self):
"""
It creates the mask from the contour and returns it.
"""
r = self.bbox[3]
c = self.bbox[2]
mask = np.zeros((r, c))
# main polygon
[rr, cc] = polygon(self.contour[:, 1], self.contour[:, 0])
rr = rr - int(self.bbox[0])
cc = cc - int(self.bbox[1])
mask[rr, cc] = 1
# holes
for inner_contour in self.inner_contours:
[rr, cc] = polygon(inner_contour[:, 1], inner_contour[:, 0])
rr = rr - int(self.bbox[0])
cc = cc - int(self.bbox[1])
mask[rr, cc] = 0
return mask
def updateUsingMask(self, new_bbox, new_mask):
self.bbox = new_bbox
self.createContourFromMask(new_mask)
self.setupForDrawing()
self.calculatePerimeter()
self.calculateArea(new_mask)
self.calculateCentroid(new_mask)
def snapToBorder(self, points):
"""
Given a curve specified as a set of points, snap the curve on the blob mask:
1) the initial segments of the curve are removed until they snap
2) the end segments of the curve are removed until they snap
"""
contour = self.contour.copy()
test = points_in_poly(points, contour)
jump = np.gradient(test.astype(int))
ind = np.nonzero(jump)
ind = np.asarray(ind)
snappoints = None
if ind.shape[1] > 2:
first_el = ind[0, 0] + 1
last_el = ind[0, -1]
snappoints = points[first_el:last_el, :].copy()
return snappoints
def createFromClosedCurve(self, points):
"""
It creates a blob starting from a closed curve. If the curve is not closed False is returned.
If the curve intersect itself many times the first segmented region is created.
"""
pt_min = np.amin(points, axis=0)
xmin = pt_min[0]
ymin = pt_min[1]
pt_max = np.amax(points, axis=0)
xmax = pt_max[0]
ymax = pt_max[1]
x_left = int(xmin) - 8
y_top = int(ymin) - 8
x_right = int(xmax) + 8
y_bottom = int(ymax) + 8
bbox = np.array([y_top, x_left, x_right - x_left, y_bottom - y_top])
mask = np.zeros((bbox[3], bbox[2]))
mask_area = bbox[3] * bbox[2]
for i in range(points.shape[0]):
x = points[i, 0]
y = points[i, 1]
yy = int(y) - bbox[0]
xx = int(x) - bbox[1]
for offsetx in range(-1, 2):
for offsety in range(-1, 2):
mask[yy + offsety, xx + offsetx] = 1
mask_flood = flood(mask, (4, 4), tolerance=0).astype(int)
for i in range(points.shape[0]):
x = points[i, 0]
y = points[i, 1]
yy = int(y) - bbox[0]
xx = int(x) - bbox[1]
for offsetx in range(-1, 2):
for offsety in range(-1, 2):
mask_flood[yy + offsety, xx + offsetx] = 1
for y in range(mask_flood.shape[0]):
for x in range(mask_flood.shape[1]):
if mask_flood[y, x] == 1:
mask[y, x] = 0
else:
mask[y, x] = 1
mask = ndi.binary_fill_holes(mask).astype(int)
# check
regions = measure.regionprops(mask)
if len(regions) != 1:
return False
else:
region = regions[0]
check = region.area / mask_area
if check > 0.95:
return False
else:
self.updateUsingMask(bbox, mask)
return True
def createCrack(self, input_arr, x, y, tolerance, preview=True):
"""
Given a inner blob point (x,y), the function use it as a seed for a paint butcket tool and create
a correspondent blob hole
"""
x_crop = x - self.bbox[1]
y_crop = y - self.bbox[0]
input_arr = gaussian(input_arr, 2)
# input_arr = segmentation.inverse_gaussian_gradient(input_arr, alpha=1, sigma=1)
blob_mask = self.getMask()
crack_mask = flood(input_arr, (int(y_crop), int(x_crop)),
tolerance=tolerance).astype(int)
cracked_blob = np.logical_and((blob_mask > 0), (crack_mask < 1))
cracked_blob = cracked_blob.astype(int)
if preview:
return cracked_blob
else:
self.updateUsingMask(self.bbox, cracked_blob)
return cracked_blob
def addToMask(self, points):
"""
Given a curve specified as a set of points, the pixels OUTSIDE the blob but inside the handle created
by the curve are added to the blob.
"""
# store the original inner contours (i.e. the holes)
original_inner_contours = []
for inner_contour in self.inner_contours:
duplicate_inner_contour = inner_contour.copy()
original_inner_contours.append(duplicate_inner_contour)
# enlarge the mask
y1A = self.bbox[0]
x1A = self.bbox[1]
x2A = x1A + self.bbox[2]
y2A = y1A + self.bbox[3]
pt_min = np.amin(points, axis=0)
xmin = pt_min[0]
ymin = pt_min[1]
pt_max = np.amax(points, axis=0)
xmax = pt_max[0]
ymax = pt_max[1]
x1B = int(xmin)
y1B = int(ymin)
x2B = int(xmax)
y2B = int(ymax)
x_left = min(x1A, x1B) - 2
y_top = min(y1A, y1B) - 2
x_right = max(x2A, x2B) + 2
y_bottom = max(y2A, y2B) + 2
bbox_union = np.array(
[y_top, x_left, x_right - x_left, y_bottom - y_top])
mask_union = np.zeros((bbox_union[3], bbox_union[2]))
blob_mask = self.getMask()
for y in range(blob_mask.shape[0]):
for x in range(blob_mask.shape[1]):
yy = y + (self.bbox[0] - bbox_union[0])
xx = x + (self.bbox[1] - bbox_union[1])
mask_union[yy, xx] = blob_mask[y, x]
for i in range(points.shape[0]):
x = points[i, 0]
y = points[i, 1]
yy = int(y) - bbox_union[0]
xx = int(x) - bbox_union[1]
for offsetx in range(-1, 2):
for offsety in range(-1, 2):
mask_union[yy + offsety, xx + offsetx] = 1
mask_union = ndi.binary_fill_holes(mask_union).astype(int)
self.updateUsingMask(bbox_union, mask_union)
# RE-ADD THE ORIGINAL INNER CONTOURS (I.E. THE HOLES)
if original_inner_contours:
self.inner_contours.clear()
for inner_contour in original_inner_contours:
# recover inner contour list
self.inner_contours.append(inner_contour)
# update qpainterpath
self.setupForDrawing()
def cutFromMask(self, points):
"""
Given a curve specified as a set of points, the pixels INSIDE the blob but "cutted" by the curve
are removed from the blob.
"""
# enlarge the mask
y1A = self.bbox[0]
x1A = self.bbox[1]
x2A = x1A + self.bbox[2]
y2A = y1A + self.bbox[3]
pt_min = np.amin(points, axis=0)
xmin = pt_min[0]
ymin = pt_min[1]
pt_max = np.amax(points, axis=0)
xmax = pt_max[0]
ymax = pt_max[1]
x1B = int(xmin)
y1B = int(ymin)
x2B = int(xmax)
y2B = int(ymax)
x_left = min(x1A, x1B) - 2
y_top = min(y1A, y1B) - 2
x_right = max(x2A, x2B) + 2
y_bottom = max(y2A, y2B) + 2
bbox_union = np.array(
[y_top, x_left, x_right - x_left, y_bottom - y_top])
mask_union = np.zeros((bbox_union[3], bbox_union[2]))
blob_mask = self.getMask()
for y in range(blob_mask.shape[0]):
for x in range(blob_mask.shape[1]):
yy = y + (self.bbox[0] - bbox_union[0])
xx = x + (self.bbox[1] - bbox_union[1])
mask_union[yy, xx] = blob_mask[y, x]
for i in range(points.shape[0]):
x = points[i, 0]
y = points[i, 1]
yy = int(y) - bbox_union[0]
xx = int(x) - bbox_union[1]
for offsetx in range(-1, 2):
for offsety in range(-1, 2):
mask_union[yy + offsety, xx + offsetx] = 0
label_image = measure.label(mask_union)
regions = measure.regionprops(label_image)
if len(regions) > 1:
# TENERE SOLO QUELLA CON AREA MASSIMA ??
area_max = 0
region_to_remove = None
for region in regions:
if region.area > area_max:
area_max = region.area
for region in regions:
if region.area < area_max:
for coords in region.coords:
mask_union[coords[0], coords[1]] = 0
self.updateUsingMask(bbox_union, mask_union)
def createContourFromMask(self, mask):
"""
It creates the contour (and the corrisponding polygon) from the blob mask.
"""
# NOTE: The mask is expected to be cropped around its bbox (!!) (see the __init__)
self.inner_contours.clear()
# we need to pad the mask to avoid to break the contour that touchs the borders
PADDED_SIZE = 2
img_padded = pad(mask, (PADDED_SIZE, PADDED_SIZE),
mode="constant",
constant_values=(0, 0))
contours = measure.find_contours(img_padded, 0.5)
number_of_contours = len(contours)
if number_of_contours > 1:
# search the longest contour
npoints_max = 0
index = 0
for i, contour in enumerate(contours):
npoints = contour.shape[0]
if npoints > npoints_max:
npoints_max = npoints
index = i
# divide the contours in OUTER contour and INNER contours
for i, contour in enumerate(contours):
if i == index:
self.contour = np.array(contour)
else:
coordinates = np.array(contour)
self.inner_contours.append(coordinates)
# adjust the coordinates of the outer contour
# (NOTE THAT THE COORDINATES OF THE BBOX ARE IN THE GLOBAL MAP COORDINATES SYSTEM)
for i in range(self.contour.shape[0]):
ycoor = self.contour[i, 0]
xcoor = self.contour[i, 1]
self.contour[i, 0] = xcoor - PADDED_SIZE + self.bbox[1]
self.contour[i, 1] = ycoor - PADDED_SIZE + self.bbox[0]
# adjust coordinates of the INNER contours
for j, contour in enumerate(self.inner_contours):
for i in range(contour.shape[0]):
ycoor = contour[i, 0]
xcoor = contour[i, 1]
self.inner_contours[j][
i, 0] = xcoor - PADDED_SIZE + self.bbox[1]
self.inner_contours[j][
i, 1] = ycoor - PADDED_SIZE + self.bbox[0]
elif number_of_contours == 1:
coords = measure.approximate_polygon(contours[0], tolerance=1.2)
self.contour = np.array(coords)
# adjust the coordinates of the outer contour
# (NOTE THAT THE COORDINATES OF THE BBOX ARE IN THE GLOBAL MAP COORDINATES SYSTEM)
for i in range(self.contour.shape[0]):
ycoor = self.contour[i, 0]
xcoor = self.contour[i, 1]
self.contour[i, 0] = xcoor - PADDED_SIZE + self.bbox[1]
self.contour[i, 1] = ycoor - PADDED_SIZE + self.bbox[0]
else:
print("ZERO CONTOURS -> THERE ARE SOME PROBLEMS HERE !!!!!!)")
def setupForDrawing(self):
"""
Create the QPolygon and the QPainterPath according to the blob's contours.
"""
# QPolygon to draw the blob
qpolygon = QPolygonF()
for i in range(self.contour.shape[0]):
qpolygon << QPointF(self.contour[i, 0], self.contour[i, 1])
self.qpath = QPainterPath()
self.qpath.addPolygon(qpolygon)
for inner_contour in self.inner_contours:
qpoly_inner = QPolygonF()
for i in range(inner_contour.shape[0]):
qpoly_inner << QPointF(inner_contour[i, 0], inner_contour[i,
1])
path_inner = QPainterPath()
path_inner.addPolygon(qpoly_inner)
self.qpath = self.qpath.subtracted(path_inner)
def createQPixmapFromMask(self):
w = self.bbox[2]
h = self.bbox[3]
self.qimg_mask = QImage(w, h, QImage.Format_ARGB32)
self.qimg_mask.fill(qRgba(0, 0, 0, 0))
if self.class_name == "Empty":
rgba = qRgba(255, 255, 255, 255)
else:
rgba = qRgba(self.class_color[0], self.class_color[1],
self.class_color[2], 100)
blob_mask = self.getMask()
for x in range(w):
for y in range(h):
if mask[y, x] == 1:
self.qimg_mask.setPixel(x, y, rgba)
self.pxmap_mask = QPixmap.fromImage(self.qimg_mask)
def calculateCentroid(self, mask):
sumx = 0.0
sumy = 0.0
n = 0
for y in range(mask.shape[0]):
for x in range(mask.shape[1]):
if mask[y, x] == 1:
sumx += float(x)
sumy += float(y)
n += 1
# NOTE: centroid is (x,y), bbox is [width height]
self.centroid[0] = int(sumx / n) + self.bbox[1]
self.centroid[1] = int(sumy / n) + self.bbox[0]
xc = int(self.centroid[0])
yc = int(self.centroid[1])
self.instance_name = "coral-" + str(xc) + "-" + str(yc)
def calculateContourPerimeter(self, contour):
# perimeter of the outer contour
px1 = contour[0, 0]
py1 = contour[0, 1]
N = contour.shape[0]
pxlast = contour[N - 1, 0]
pylast = contour[N - 1, 1]
perim = math.sqrt((px1 - pxlast) * (px1 - pxlast) + (py1 - pylast) *
(py1 - pylast))
for i in range(1, contour.shape[0]):
px2 = contour[i, 0]
py2 = contour[i, 1]
d = math.sqrt((px1 - px2) * (px1 - px2) + (py1 - py2) *
(py1 - py2))
perim += d
px1 = px2
py1 = py2
return perim
def calculatePerimeter(self):
self.perimeter = self.calculateContourPerimeter(self.contour)
for contour in self.inner_contours:
self.perimeter += self.calculateContourPerimeter(self.contour)
def calculateArea(self, mask):
self.area = 0.0
for y in range(mask.shape[0]):
for x in range(mask.shape[1]):
if mask[y, x] == 1:
self.area += 1.0
def fromDict(self, dict):
"""
Set the blob information given it represented as a dictionary.
"""
self.bbox = np.asarray(dict["bbox"])
self.centroid = np.asarray(dict["centroid"])
self.area = dict["area"]
self.perimeter = dict["perimeter"]
self.contour = np.asarray(dict["contour"])
inner_contours = dict["inner contours"]
self.inner_contours = []
for c in inner_contours:
self.inner_contours.append(np.asarray(c))
self.deep_extreme_points = np.asarray(dict["deep_extreme_points"])
self.class_name = dict["class name"]
self.class_color = dict["class color"]
self.instance_name = dict["instance name"]
self.blob_name = dict["blob name"]
self.id = dict["id"]
self.note = dict["note"]
# finalize blob
self.setupForDrawing()
def toDict(self):
"""
Get the blob information as a dictionary.
"""
dict = {}
dict["bbox"] = self.bbox.tolist()
dict["centroid"] = self.centroid.tolist()
dict["area"] = self.area
dict["perimeter"] = self.perimeter
dict["contour"] = self.contour.tolist()
dict["inner contours"] = []
for c in self.inner_contours:
dict["inner contours"].append(c.tolist())
dict["deep_extreme_points"] = self.deep_extreme_points.tolist()
dict["class name"] = self.class_name
dict["class color"] = self.class_color
dict["instance name"] = self.instance_name
dict["blob name"] = self.blob_name
dict["id"] = self.id
dict["note"] = self.note
return dict
class Blob(object):
"""
Blob data. A blob is a group of pixels.
A blob can be tagged with the class and other information.
Both the set of pixels and the corresponding vectorized version are stored.
"""
def __init__(self, region, offset_x, offset_y, id):
if region == None: # AN EMPTY BLOB IS CREATED..
self.area = 0.0
self.perimeter = 0.0
self.centroid = np.zeros((2))
self.bbox = np.zeros((4))
# placeholder; empty contour
self.contour = np.zeros((2, 2))
self.inner_contours = []
self.qpath = None
self.qpath_gitem = None
self.instance_name = "noname"
self.blob_name = "noname"
self.id = 0
else:
# extract properties
self.centroid = np.array(region.centroid)
cy = self.centroid[0]
cx = self.centroid[1]
self.centroid[0] = cx + offset_x
self.centroid[1] = cy + offset_y
# Bounding box (min_row, min_col, max_row, max_col).
# Pixels belonging to the bounding box are in the half-open
# interval [min_row, max_row) and [min_col, max_col).
self.bbox = np.array(region.bbox)
width = self.bbox[3] - self.bbox[1]
height = self.bbox[2] - self.bbox[0]
# BBOX -> TOP, LEFT, WIDTH, HEIGHT
self.bbox[0] = self.bbox[0] + offset_y
self.bbox[1] = self.bbox[1] + offset_x
self.bbox[2] = width
self.bbox[3] = height
# QPainterPath associated with the contours
self.qpath = None
# QGraphicsItem associated with the QPainterPath
self.qpath_gitem = None
# to extract the contour we use the mask cropped according to the bbox
input_mask = region.image.astype(int)
self.contour = np.zeros((2, 2))
self.inner_contours = []
self.updateUsingMask(self.bbox, input_mask)
# a string with a progressive number to identify the instance
self.instance_name = "coral" + str(id)
# a string with a number to identify the blob plus its centroid
xc = int(self.centroid[0])
yc = int(self.centroid[1])
self.blob_name = "blob" + str(id) + "-" + str(xc) + "-" + str(yc)
self.id = id
# deep extreme points (for fine-tuning)
self.deep_extreme_points = np.zeros((4, 2))
# name of the class
self.class_name = "Empty"
# color of the class
self.class_color = [128, 128, 128]
# note about the coral, i.e. damage type
self.note = ""
# QImage corresponding to the current mask
self.qimg_mask = None
# QPixmap associated with the mask (for pixel-level editing operations)
self.pxmap_mask = None
# QGraphicsItem associated with the pixmap mask
self.pxmap_mask_gitem = None
# membership group (if any)
self.group = None
def copy(self):
blob = Blob(None, 0, 0, 0)
blob.area = self.area
blob.perimeter = self.perimeter
blob.centroid = self.centroid
blob.bbox = self.bbox
blob.contour = self.contour.copy()
for inner in self.inner_contours:
blob.inner_contours.append(inner.copy())
blob.qpath_gitem = None
blob.qpath = None
blob.instance_name = blob.instance_name
blob.blob_name = self.blob_name
blob.id = self.id
blob.class_name = self.class_name
blob.deep_extreme_points = self.deep_extreme_points
self.note = ""
self.qimg_mask = None
self.pxmap_mask = None
self.pxmap_mask_gitem = None
return blob
def __deepcopy__(self, memo):
#avoid recursion!
deepcopy_method = self.__deepcopy__
self.__deepcopy__ = None
#save and later restore qobjects
path = self.qpath
pathitem = self.qpath_gitem
#no deep copy for qobjects
self.qpath = None
self.qpath_gitem = None
blob = copy.deepcopy(self)
blob.contour = self.contour.copy()
blob.inner_contours.clear()
for inner in self.inner_contours:
blob.inner_contours.append(inner.copy())
blob.qpath = None
blob.qpath_gitem = None
self.qpath = path
self.qpath_gitem = pathitem
#restore deepcopy (also to the newly created Blob!
blob.__deepcopy__ = self.__deepcopy__ = deepcopy_method
return blob
def setId(self, id):
# a string with a number to identify the blob plus its centroid
xc = int(self.centroid[0])
yc = int(self.centroid[1])
self.blob_name = "blob" + str(id) + "-" + str(xc) + "-" + str(yc)
self.id = id
def getMask(self):
"""
It creates the mask from the contour and returns it.
"""
r = self.bbox[3]
c = self.bbox[2]
mask = np.zeros((r, c), dtype=np.bool_)
# main polygon
[rr, cc] = polygon(self.contour[:, 1], self.contour[:, 0])
rr = rr - int(self.bbox[0])
cc = cc - int(self.bbox[1])
mask[rr, cc] = 1
# holes
for inner_contour in self.inner_contours:
[rr, cc] = polygon(inner_contour[:, 1], inner_contour[:, 0])
rr = rr - int(self.bbox[0])
cc = cc - int(self.bbox[1])
mask[rr, cc] = 0
return mask
def updateUsingMask(self, bbox, mask):
self.bbox = bbox
self.createContourFromMask(mask)
self.calculatePerimeter()
self.calculateCentroid(mask)
self.calculateArea(mask)
def lineToPoints(self, lines, snap=False):
points = np.empty(shape=(0, 2), dtype=int)
for line in lines:
p = self.drawLine(line)
if p.shape[0] == 0:
continue
if snap:
p = self.snapToBorder(p)
if p is None:
continue
points = np.append(points, p, axis=0)
return points
def drawLine(self, line):
(x, y) = utils.draw_open_polygon(line[:, 1], line[:, 0])
points = np.asarray([x, y]).astype(int)
points = points.transpose()
points[:, [1, 0]] = points[:, [0, 1]]
return points
def snapToBorder(self, points):
return self.snapToContour(points, self.contour)
def snapToContour(self, points, contour):
"""
Given a curve specified as a set of points, snap the curve on the blob mask:
1) the initial segments of the curve are removed until they snap
2) the end segments of the curve are removed until they snap
"""
test = points_in_poly(points, contour)
if test is None or test.shape[0] == 0:
return None
jump = np.gradient(test.astype(int))
ind = np.nonzero(jump)
ind = np.asarray(ind)
snappoints = None
if ind.shape[1] > 2:
first_el = ind[0, 0] + 1
last_el = ind[0, -1]
snappoints = points[first_el:last_el, :].copy()
return snappoints
def snapToInternalBorders(self, points):
if not self.inner_contours:
return None
snappoints = np.zeros(shape=(0, 2))
for contour in self.inner_contours:
snappoints = np.append(snappoints,
self.snapToContour(points, contour))
return snappoints
def createFromClosedCurve(self, lines):
"""
It creates a blob starting from a closed curve. If the curve is not closed False is returned.
If the curve intersect itself many times the first segmented region is created.
"""
points = self.lineToPoints(lines)
box = Mask.pointsBox(points, 4)
(mask, box) = Mask.jointMask(box, box)
Mask.paintPoints(mask, box, points, 1)
mask = ndi.binary_fill_holes(mask)
mask = binary_erosion(mask)
mask = binary_erosion(mask)
mask = binary_erosion(mask)
mask = binary_dilation(mask)
mask = binary_dilation(mask)
self.updateUsingMask(box, mask)
return True
def createCrack(self, input_arr, x, y, tolerance, preview=True):
"""
Given a inner blob point (x,y), the function use it as a seed for a paint butcket tool and create
a correspondent blob hole
"""
x_crop = x - self.bbox[1]
y_crop = y - self.bbox[0]
input_arr = gaussian(input_arr, 2)
# input_arr = segmentation.inverse_gaussian_gradient(input_arr, alpha=1, sigma=1)
blob_mask = self.getMask()
crack_mask = flood(input_arr, (int(y_crop), int(x_crop)),
tolerance=tolerance).astype(int)
cracked_blob = np.logical_and((blob_mask > 0), (crack_mask < 1))
cracked_blob = cracked_blob.astype(int)
if not preview:
self.updateUsingMask(self.bbox, cracked_blob)
return cracked_blob
def createContourFromMask(self, mask):
"""
It creates the contour (and the corrisponding polygon) from the blob mask.
"""
# NOTE: The mask is expected to be cropped around its bbox (!!) (see the __init__)
self.inner_contours.clear()
# we need to pad the mask to avoid to break the contour that touches the borders
PADDED_SIZE = 2
img_padded = pad(mask, (PADDED_SIZE, PADDED_SIZE),
mode="constant",
constant_values=(0, 0))
contours = measure.find_contours(img_padded, 0.5)
number_of_contours = len(contours)
threshold = 20 #min number of points in a small hole
if number_of_contours > 1:
# search the longest contour
npoints_max = 0
index = 0
for i, contour in enumerate(contours):
npoints = contour.shape[0]
if npoints > npoints_max:
npoints_max = npoints
index = i
# divide the contours in OUTER contour and INNER contours
for i, contour in enumerate(contours):
if i == index:
self.contour = np.array(contour)
else:
if contour.shape[0] > threshold:
coordinates = np.array(contour)
self.inner_contours.append(coordinates)
# adjust the coordinates of the outer contour
# (NOTE THAT THE COORDINATES OF THE BBOX ARE IN THE GLOBAL MAP COORDINATES SYSTEM)
for i in range(self.contour.shape[0]):
ycoor = self.contour[i, 0]
xcoor = self.contour[i, 1]
self.contour[i, 0] = xcoor - PADDED_SIZE + self.bbox[1]
self.contour[i, 1] = ycoor - PADDED_SIZE + self.bbox[0]
# adjust coordinates of the INNER contours
for j, contour in enumerate(self.inner_contours):
for i in range(contour.shape[0]):
ycoor = contour[i, 0]
xcoor = contour[i, 1]
self.inner_contours[j][
i, 0] = xcoor - PADDED_SIZE + self.bbox[1]
self.inner_contours[j][
i, 1] = ycoor - PADDED_SIZE + self.bbox[0]
elif number_of_contours == 1:
coords = measure.approximate_polygon(contours[0], tolerance=1.2)
self.contour = np.array(coords)
# adjust the coordinates of the outer contour
# (NOTE THAT THE COORDINATES OF THE BBOX ARE IN THE GLOBAL MAP COORDINATES SYSTEM)
for i in range(self.contour.shape[0]):
ycoor = self.contour[i, 0]
xcoor = self.contour[i, 1]
self.contour[i, 0] = xcoor - PADDED_SIZE + self.bbox[1]
self.contour[i, 1] = ycoor - PADDED_SIZE + self.bbox[0]
else:
print("ZERO CONTOURS -> THERE ARE SOME PROBLEMS HERE !!!!!!)")
def setupForDrawing(self):
"""
Create the QPolygon and the QPainterPath according to the blob's contours.
"""
# QPolygon to draw the blob
qpolygon = QPolygonF()
for i in range(self.contour.shape[0]):
qpolygon << QPointF(self.contour[i, 0], self.contour[i, 1])
self.qpath = QPainterPath()
self.qpath.addPolygon(qpolygon)
for inner_contour in self.inner_contours:
qpoly_inner = QPolygonF()
for i in range(inner_contour.shape[0]):
qpoly_inner << QPointF(inner_contour[i, 0], inner_contour[i,
1])
path_inner = QPainterPath()
path_inner.addPolygon(qpoly_inner)
self.qpath = self.qpath.subtracted(path_inner)
def createQPixmapFromMask(self):
w = self.bbox[2]
h = self.bbox[3]
self.qimg_mask = QImage(w, h, QImage.Format_ARGB32)
self.qimg_mask.fill(qRgba(0, 0, 0, 0))
if self.class_name == "Empty":
rgba = qRgba(255, 255, 255, 255)
else:
rgba = qRgba(self.class_color[0], self.class_color[1],
self.class_color[2], 100)
blob_mask = self.getMask()
for x in range(w):
for y in range(h):
if blob_mask[y, x] == 1:
self.qimg_mask.setPixel(x, y, rgba)
self.pxmap_mask = QPixmap.fromImage(self.qimg_mask)
def calculateCentroid(self, mask):
m = measure.moments(mask)
c = np.array((m[0, 1] / m[0, 0], m[1, 0] / m[0, 0]))
#centroid is (x, y) while measure returns (y,x and bbox is yx)
self.centroid = np.array((c[1] + self.bbox[1], c[0] + self.bbox[0]))
self.blob_name = "coral-" + str(self.centroid[0]) + "-" + str(
self.centroid[1])
def calculateContourPerimeter(self, contour):
#self.perimeter = measure.perimeter(mask) instead?
# perimeter of the outer contour
px1 = contour[0, 0]
py1 = contour[0, 1]
N = contour.shape[0]
pxlast = contour[N - 1, 0]
pylast = contour[N - 1, 1]
perim = math.sqrt((px1 - pxlast) * (px1 - pxlast) + (py1 - pylast) *
(py1 - pylast))
for i in range(1, contour.shape[0]):
px2 = contour[i, 0]
py2 = contour[i, 1]
d = math.sqrt((px1 - px2) * (px1 - px2) + (py1 - py2) *
(py1 - py2))
perim += d
px1 = px2
py1 = py2
return perim
def calculatePerimeter(self):
self.perimeter = self.calculateContourPerimeter(self.contour)
for contour in self.inner_contours:
self.perimeter += self.calculateContourPerimeter(self.contour)
def calculateArea(self, mask):
self.area = mask.sum().astype(float)
def fromDict(self, dict):
"""
Set the blob information given it represented as a dictionary.
"""
self.bbox = np.asarray(dict["bbox"])
self.centroid = np.asarray(dict["centroid"])
self.area = dict["area"]
self.perimeter = dict["perimeter"]
self.contour = np.asarray(dict["contour"])
inner_contours = dict["inner contours"]
self.inner_contours = []
for c in inner_contours:
self.inner_contours.append(np.asarray(c))
self.deep_extreme_points = np.asarray(dict["deep_extreme_points"])
self.class_name = dict["class name"]
self.class_color = dict["class color"]
self.instance_name = dict["instance name"]
self.blob_name = dict["blob name"]
self.id = dict["id"]
self.note = dict["note"]
# finalize blob
#self.setupForDrawing()
def toDict(self):
"""
Get the blob information as a dictionary.
"""
dict = {}
dict["bbox"] = self.bbox.tolist()
dict["centroid"] = self.centroid.tolist()
dict["area"] = self.area
dict["perimeter"] = self.perimeter
dict["contour"] = self.contour.tolist()
dict["inner contours"] = []
for c in self.inner_contours:
dict["inner contours"].append(c.tolist())
dict["deep_extreme_points"] = self.deep_extreme_points.tolist()
dict["class name"] = self.class_name
dict["class color"] = self.class_color
dict["instance name"] = self.instance_name
dict["blob name"] = self.blob_name
dict["id"] = self.id
dict["note"] = self.note
return dict
