def find_blobs(image,size):
center_x = size[0]/2
center_y = size[1]/2
params = cv2.SimpleBlobDetector_Params()
params.filterByCircularity = True
params.minCircularity = 0
params.maxCircularity = 1
params.minDistBetweenBlobs = 5
params.filterByArea = True
params.minArea = 1
params.maxArea = 500000
params.filterByInertia = True
params.minInertiaRatio = 0
params.maxInertiaRatio = 1
params.filterByConvexity = True
params.minConvexity = 0
params.maxConvexity = 1
detector = cv2.SimpleBlobDetector_create(params)
found_blobs = detector.detect(image)
blobs = []
for blob in found_blobs:
blob = Blob(blob.pt[0]-20, blob.pt[1]-20, blob.size/2)
blob.set_center(center_x, center_y)
if not is_score(blob, size):
blobs.append(blob)
#image = cv2.drawKeypoints(image, found_blobs, np.array([]), (0,0,255), cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
#cv2.imshow("Keypoints", image)
#cv2.waitKey(0)
return blobs
def _apply_fun(self, blob_generator, fun):
threadsafe_generator = ThreadsafeIter(blob_generator,len(self.pipelines))
executor = concurrent.futures.ThreadPoolExecutor(max_workers=len(self.pipelines))
# Start the load operations and mark each future with its URL
futures = [executor.submit(fun, pipeline, threadsafe_generator) for pipeline in self.pipelines]
generators = [gen for gen in concurrent.futures.as_completed(futures)]
has_more_blobs = True
while has_more_blobs:
b = Blob()
out_blobs = [next(gen.result()) for gen in generators]
blob_uuids = [blob.meta.uuid for blob in out_blobs]
# If not all UUIDs are equal or num outputs is not the same as the number of pipelines
if not out_blobs or blob_uuids.count(blob_uuids[0]) != len(blob_uuids) or len(out_blobs) != len(self.pipelines):
logging.error("Number of elements changed within ParallelAlgorithm pipelines. This is not allowed!")
raise Exception("Error")
# If there are no more blobs, we are done
if len(out_blobs) == 0:
has_more_blobs = False
else:
b.data = [blob.data.ravel() for blob in out_blobs]
b.data = hstack(b.data)
b.meta = out_blobs[0].meta
yield b
logging.info("Finished training in ParallelAlgorithm")
def test_blob(image):
regions = Region.find_regions_by_border_color(image)
for i, region in enumerate(regions):
region_mask = region.mask_image(image)
blobs = region.find_blobs(image)
Blob.draw_outlines(blobs, region_mask, (234, 34, 102))
utils.show_image(region_mask, time=1000)
def rectangle(x0,y0,x1,y1,res,math):
'''Create a rectangular blob of charge given coordinates which represent
diagonal corners.
Params:
res -> Correlated with the 'resolution' of the rectangle. Larger
will lead to longer and finer calculations
math -> Math is the string containing a Polish notation function
of X, Y (TODO: add T). It is parsed into a function
using mathparse.math_parse'''
rect = Blob() ## Initialize a new Blob...
x0, y0, x1, y1 = float(x0), float(y0), float(x1), float(y1)
xs = linspace(x0,x1, res/(y1-y0)) ## Set up a vertical and horizontal line
ys = linspace(y0,y1, res/(x1-x0)) ## of charges
for x in xs: ## Basically set up a meshgrid of x and y
for y in ys: ## TODO: Look into replacing with Meshgrid.
rect.add_point(Point(x,y)) ## Add each point to the blob
f = math_parse(str(math)) ## Parse the Polish notation into python
## syntax.
rect.math = lambda x, y: eval(f) ## Evaluate the python syntax string
## into an expression, and bind it to
## a function at the math method of
## our blob
blobs.append(rect) ## Put it on the list for later flattening
def blob_detector2(image, stepSize, windowSize):
blobs = []
print("loading model...")
model = load_model(
'/Users/2020shatgiskessell/Desktop/New_Mole_Detector/Mole_Detector_1_3/my_mole_model_2.h5'
)
for y in range(0, image.shape[0], stepSize):
for x in range(0, image.shape[1], stepSize):
# identify moles on current window
#blobs.append(identify_moles(image[y:y + windowSize[1], x:x + windowSize[0]]))
roi = image[y:y + windowSize[1], x:x + windowSize[0]]
try:
roi = cv2.resize(roi, (8, 8))
except Exception:
continue
roi = np.expand_dims(roi, axis=2)
roi = np.expand_dims(roi, axis=0)
pred = model.predict(roi)
pred = pred.round()
if int(pred[0][0]) == 1:
blob = Blob()
blob.x = x
blob.y = y
#blob.matrix = component_matrix
blob.roi = roi
blobs.append(blob)
return blobs
def parametric_curve(s1, s2, f1, f2, res, math):
''' Creates a blob which is a parametric curve of the form f(x,y) = (g(t),h(t))
s1, s2 donate the domain of the parameter, fx is the x component, fy is the y compenent
'''
curve = Blob() #New blob
s1,s2 = float(s1), float(s2)
domain = linspace(s1,s2,res) # Domain of the parameterizing variable
f1_func = lambda s: eval(math_parse(f1)) # We're gonna have a two part math here, in
f2_func = lambda s: eval(math_parse(f2)) # addition to the usual to calculate x and y
calc_charges = lambda x, y: eval (math_parse(str(math)))
xs = f1_func(domain)
ys = f2_func(domain)
for x,y in zip(xs, ys):
curve.add_point(Point(x,y))
curve.math = calc_charges
blobs.append(curve)
def compute(self, image):
b = Blob()
if isinstance(image, str):
b.meta.imagepath = image
else:
b.data = image
# Use compute all here to incorporate custom compute all functions
return next(self.pipeline.compute([b])).data
def _train(self, blob_generator):
# First, collect all elements of the input
data = []
labels = []
metas = []
for blob in blob_generator:
if self.use_sparse is None:
# Determine automatically by comparing size
sparse_vec = scipy.sparse.csr_matrix(blob.data.ravel())
sparse_memory_req = sparse_vec.data.nbytes + sparse_vec.indptr.nbytes + sparse_vec.indices.nbytes
self.use_sparse = sparse_memory_req < blob.data.nbytes
logging.debug(
'Using sparse format for collecting features: %s' %
self.use_sparse)
logging.debug('Blob data needs %i' % blob.data.nbytes)
logging.debug('%i with sparse vs %i with dense' %
(sparse_memory_req, blob.data.nbytes))
if self.use_sparse:
data.append(scipy.sparse.csr_matrix(blob.data.ravel()))
else:
data.append(blob.data.ravel())
labels.append(blob.meta.label)
metas.append(blob.meta)
# Stack data to matrix explicitly here, as both fit and predict
# would to this stacking otherwise
try:
if self.use_sparse:
data = scipy.sparse.vstack(data)
data = data.astype(np.float64)
else:
data = np.array(data, dtype=np.float64)
except ValueError:
logging.error(
"Length of all feature vectors need to be the same for Classificator training."
)
raise Exception
logging.warning(
'Training the model with feature dim %i, this might take a while' %
data.shape[1])
self.model.fit(data, labels)
logging.warning('Finished')
for (d, m) in zip(self.model.decision_function(data), metas):
b = Blob()
b.data = d
b.meta = m
yield b
def _gen_inblobs(self, images, prog_bar=True):
if prog_bar:
im_gen = pyprind.prog_bar(list(images))
else:
im_gen = list(images)
for im in im_gen:
if isinstance(im, Blob):
b = im
else:
b = Blob()
if isinstance(im, str):
b.meta.imagepath = im
else:
b.data = im
yield b
def circle(x0,y0,radius,start_arc,end_arc,numpoints,math):
"""Create a circle of charge by using polar notation. Circle will have a start and stop end_arc
given in degrees (for covinience)."""
circle = Blob() ##Spiffy new blob
x0, y0, start_arc, end_arc = float(x0), float(y0), float(start_arc), float(end_arc)
theta = radians(linspace(start_arc,end_arc,numpoints)) #figure out our angles
xs = radius*cos(theta)+x0 #Build up our xs and ys
ys = radius*sin(theta)+y0
for x,y in zip(xs,ys):
circle.add_point(Point(x,y))
f = math_parse(str(math)) #Parse some math for the charges
circle.math = lambda x, y: eval(f)
blobs.append(circle) #Stick into the main blob
def line(x0,y0,x1,y1,res,math):
'''Create a linear blob given start and stop corridnates, a number of points
on the line, and a math expression to be parsed by mathparse.math_parse.'''
lin = Blob() ## Iniitialize a new blob
x0, y0, x1, y1 = float(x0), float(y0), float(x1), float(y1)
horizontal = linspace(x0,x1,res) ## Make a line along x
m = (y1-y0)/(x1-x0) ## Map horizontal onto the line
b = y0- m*(x1-x0) ## using y-mx + b
for x,y in zip(horizontal, (m*horizontal) + b):
lin.add_point(Point(x,y))
f = math_parse(str(math)) ## Parse and assign math to blob.math()
lin.math = lambda x, y: eval(f)
blobs.append(lin) ## Put it on the list for later flattening
def setup():
global blob
size(640, 480)
background(250)
no_stroke()
splat = 3
stride = 5
number_of_blobs = 1
blob = [0] * number_of_blobs
print(blob)
for n in range(len(blob)):
blob[n] = Blob(width, height, splat, stride)
def _train(self, blob_generator):
# If you need all data at once:
# Remember the metas!
# Example
data = []
labels = []
metas = []
for blob in blob_generator:
data.append(blob.data.ravel())
labels.append(blob.meta.label)
metas.append(blob.meta)
numpy_data = vstack(data)
# process numpy_data
# ...
# Create generator for next layer
for d, m in zip(data, metas):
b = Blob()
b.data = d
b.meta = m
yield b
def main():
parser = argparse.ArgumentParser(description='Test blob functionality')
parser.add_argument('--file', '-f', dest='file', default='test_images/balls.png', help='file to test', type=str)
args = parser.parse_args()
image = cv2.imread(args.file)
blobs = Blob.find_blobs(image)
for blob in blobs:
blob.draw_outline(image, color=(0, 255, 0))
blob.draw_fill(image, color=(255,0,0))
print blob
cv.NamedWindow('a_window', cv.CV_WINDOW_AUTOSIZE)
cv2.imshow('a_window', image)
cv.WaitKey(10000)
def _train(self, blob_generator):
# First, collect all elements of the input
data = []
labels = []
metas = []
for blob in blob_generator:
data.append(self._add_bias(blob.data.ravel()))
labels.append(blob.meta.label)
metas.append(blob.meta)
try:
data = vstack(data)
except ValueError:
logging.error(
"Size of all input data needs to be the same for SVM training."
)
raise Exception
self.svm_model.fit(data, labels)
for (d, m) in zip(self.svm_model.predict(data), metas):
b = Blob()
b.data = d
b.meta = m
yield b
def blob_generator(self):
if len(self.imagepaths) != len(self.labels) or len(self.labels) != len(
self.split_assignments):
logging.error(
"Size of imagepaths, labels and split assignments are not equal!"
)
raise Exception
for path, label, split_assignment, idx in zip(self.imagepaths,
self.labels,
self.split_assignments,
range(len(self.labels))):
b = Blob()
b.meta.label = label
b.meta.imagepath = path
b.meta.split_assignment = split_assignment
yield b
def find_blobs(self, image):
blobs = Blob.find_blobs(self.mask_image(image))
self.blobs = [blob for blob in blobs if blob.area < self.area*self.MAX_BLOB_RATIO]
return self.blobs
def compute_stats(num_labels, labels, stats, centroids, img, x_i, y_i, i, og):
blobs = []
#start = timeit.default_timer()
#get component centroid coordinates
x, y = centroids[i]
#compute statistics
# cv2.CC_STAT_LEFT The leftmost (x) coordinate which is the inclusive start of the bounding box in the horizontal direction.
# cv2.CC_STAT_TOP The topmost (y) coordinate which is the inclusive start of the bounding box in the vertical direction.
# cv2.CC_STAT_WIDTH The horizontal size of the bounding box
# cv2.CC_STAT_HEIGHT The vertical size of the bounding box
# cv2.CC_STAT_AREA The total area (in pixels) of the connected component
width = stats[i, cv2.CC_STAT_WIDTH]
height = stats[i, cv2.CC_STAT_HEIGHT]
if height > 20:
return None, None, None, None, None, None
radius = (height + width) / 2
area = stats[i, cv2.CC_STAT_AREA]
#these are the top left x, y coordinates = ONLY TO BE USED FOR GETTING ROI
x_l = stats[i, cv2.CC_STAT_LEFT]
y_l = stats[i, cv2.CC_STAT_TOP]
#compute line
#slope, y_int= compute_lobf(x,y,x_l,y_l)
#stop = timeit.default_timer()
#print('Time to calculate properties: ', stop - start)
#remove everything except for component i to create isolated component matrix
#get connected component roi
roi = og[y_l - 1:y_l + height + 1, x_l - 1:x_l + width + 1]
#stop = timeit.default_timer()
#print('Time to create cm and roi: ', stop - start)
#----------------MEASURES-------------------------------------------------------------------
#radius = (height + width)/2
#compute more statistics related to roundness
radius = np.sqrt((area / np.pi))
formfactor = compute_formfactor(radius, area)
bounding_box_area_ratio = area / (height * width)
if height > width:
roundness = compute_roundness(height, radius, area)
aspect_ratio = height / width
else:
roundness = compute_roundness(width, radius, area)
aspect_ratio = width / height
#print('Time to calculate heuristic properties: ', stop - start)
# if x >300 and y < 150:
# print ("(" + str(x) + ","+str(y)+") -> area ratio: " + str(area/(height*width)))
#print ("(" + str(x) + ","+str(y)+") -> " + "radius: " + str(radius) + ", formfactor: " + str(formfactor) + ", roundness: " + str(roundness) + ", aspect ratio: " + str(aspect_ratio))
#print ("(" + str(x) + ","+str(y)+")")
#print ("\n")
#calculates line of best fit and error
try:
cord1, cord2, error = compute_lobf(roi, x_l * y_l)
x1, y1 = cord1
x2, y2 = cord2
except TypeError:
print("cant calculate line of best fit")
error = 0
cv2.imwrite(
os.path.join(
"/Users/2020shatgiskessell/Desktop/New_Mole_Detector/ANN_Images_Fiverr",
"roi19" + str(i) + ".png"), roi)
#COMMENT OUT WHEN COLLECTING ANN DATA!!!!!
#if the error is below 16, (the line of best fit closely matches connect component) return none
if error < 16:
return None, None, None, None, None, None
# print ("(" + str(x1) + ","+str(y1)+")")
# print ("(" + str(x2) + ","+str(y2)+")")
# print ("\n")
#next step: calculate residuals and do some sort of analysis
#try and bounding_box_area_ratio >= 0.5
if roundness > 0.2 and 0.9 < aspect_ratio < 3 and 1.1 >= formfactor > 0.9:
blob = Blob()
blob.radius = radius
blob.x = x + x_i
blob.y = y + y_i
#blob.matrix = component_matrix
blob.roi = roi
blobs.append(blob)
return blobs
def __init__(self, x_boundary, y_boundary):
Blob.__init__(self, (0, 255, 0), x_boundary, y_boundary)
def start(self, Image, ROIs):
"""Image processing function.
\param image (the enhanced and segmented image)
\return image (the annotated image )
local variable is the list of detected blobs with the following feature columns:
[bb_left,bb_top,bb_width,bb_height, cc_area, sharpness, SNR]
Sharpness is variation of the Laplacian (introduced by Pech-Pacheco
"Diatom autofocusing in brightfield microscopy: a comparative study."
"""
try:
self.startTimer()
self.image = Image
self.ROIs = ROIs
self._Blobs = list()
# Iterate ROis
for ROI in ROIs:
# slice image, assuming ROI:(left,top,width,height)
ROI_image = self.image[ROI[1]:ROI[1] + ROI[3],
ROI[0]:ROI[0] + ROI[2]]
# Binarize and find blobs
BWImage = cv2.adaptiveThreshold(ROI_image, 255,
cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
self.invBin, self.blocksize,
self.offset)
# ConnectedComponentsWithStats output: number of labels, label matrix, stats(left,top,width,height), area
_, _, blobFeatures, _ = cv2.connectedComponentsWithStats(
BWImage, 8, cv2.CV_32S)
# Get blob RoI and area
blobFeatures = blobFeatures[
1:] # skipping background (label 0)
# Filter by blob area
blobFeatures = blobFeatures[np.where(
(blobFeatures[:, cv2.CC_STAT_AREA] > self.minBlobArea)
& (blobFeatures[:, cv2.CC_STAT_AREA] < self.maxBlobArea))]
# Increase array size
blobFeatures = np.concatenate([
blobFeatures,
np.zeros((blobFeatures.shape[0], 3), dtype=int)
],
axis=1)
# Annotate blobs and compute additional features
for blob in blobFeatures:
tl = (blob[0], blob[1])
br = (blob[0] + blob[2], blob[1] + blob[3])
# Compute some metrics of individual blobs
tempImage = self.image[tl[1]:br[1], tl[0]:br[0]]
I_0 = 255.0 - np.min(
tempImage) # peak foreground intensity estimate
I_b = 255.0 - np.max(tempImage) # background intensity
# Shift coordinates wrt ROI
blob[0] += ROI[0]
blob[1] += ROI[1]
#centroid
cX = int((blob[0] + blob[0] + blob[2]) / 2.0) # x1+x2 /2
cY = int((blob[1] - blob[3] + blob[1]) / 2.0) # y1+y2 /2
DetectedBlob = Blob(blob[0], blob[0] + blob[2],
blob[1] - blob[3], blob[1], (cX, cY))
# Local sharpness column
DetectedBlob._local_sharpness = int(
cv2.Laplacian(tempImage, cv2.CV_64F).var())
# Local SNR column
DetectedBlob._local_SNR = int(
(I_0 - I_b) / np.sqrt(I_b)) if I_b > 0 else 0
# Perimeter
tempBWImage = BWImage[tl[1]:br[1], tl[0]:br[0]]
contours, _ = cv2.findContours(tempBWImage, cv2.RETR_LIST,
cv2.CHAIN_APPROX_NONE)
contour = max(
contours,
key=cv2.contourArea) # select largest contour
DetectedBlob._perimeter = len(contour)
#Add blob to list
self._Blobs.append(DetectedBlob)
# Mark in image
# if self.plot:
# cv2.rectangle(ROI_image, tl, br, (0, 0, 0), 1)
#cv2.putText(ROI_image, str(blob[5]), br, cv2.FONT_HERSHEY_SIMPLEX, .5, (0,0,0), 1, cv2.LINE_AA)
# Plot last ROI
if self.plot:
cv2.imshow(self.name, BWImage)
# Finalize
self.stopTimer()
self.signals.finished.emit()
except Exception as err:
exc = traceback.format_exception(type(err),
err,
err.__traceback__,
chain=False)
self.signals.error.emit(exc)
self.signals.message.emit('E: {} exception: {}'.format(
self.name, err))
return self.image
def main():
filepath = videopath
print filepath
video = cv2.VideoCapture("./" + videopath)
width = video.get(cv2.CAP_PROP_FRAME_WIDTH) # float video.get(3)
height = video.get(cv2.CAP_PROP_FRAME_HEIGHT) # float video.get(4)
if not video.isOpened():
print("Video not Opened!")
return
#Maybe we should check if the video has at least 2 frames
#Read the first and second frame of the video to start doing processing on them
_, imgFrame1 = video.read()
_, imgFrame2 = video.read()
#start framecount as 2 because we just read 2 frames
atFrame = 2
#up to this point we have none blobs yet
blobs = []
# To count people and pass it as a parameter. It doens't work with primitive variables (int)
peopleCount = [0]
seenPeople = set()
#While the video is open and we don't press q key read, process and show a frame
while (video.isOpened()):
#for every frame, check how many blobs are in the screen
currentBlobs = []
imgFrame1Copy = copy.deepcopy(imgFrame1)
imgFrame2Copy = copy.deepcopy(imgFrame2)
imgFrame1Copy = cv2.cvtColor(imgFrame1Copy, cv2.COLOR_BGR2GRAY)
imgFrame2Copy = cv2.cvtColor(imgFrame2Copy, cv2.COLOR_BGR2GRAY)
if (debugGaussian and debug_mode):
cv2.imshow('gaussianBlurBefore-Img1', imgFrame1Copy)
cv2.imshow('gaussianBlurBefore-Img2', imgFrame2Copy)
imgFrame1Copy = cv2.GaussianBlur(imgFrame1Copy, gaussian_kernel, 0)
imgFrame2Copy = cv2.GaussianBlur(imgFrame2Copy, gaussian_kernel, 0)
if (debugGaussian and debug_mode):
cv2.imshow('gaussianBlurAfter-Img1', imgFrame1Copy)
cv2.imshow('gaussianBlurAfter-Img2', imgFrame2Copy)
imgDifference = cv2.absdiff(imgFrame1Copy, imgFrame2Copy)
if (debugGaussian and debug_mode):
cv2.imshow('dif-Img1-Img2', imgDifference)
# ret value is used for Otsu's Binarization if we want to
# https://docs.opencv.org/3.4.0/d7/d4d/tutorial_py_thresholding.html
ret, imgThresh = cv2.threshold(imgDifference, threshold_value, 255.0,
cv2.THRESH_BINARY)
if debugThreshold and debug_mode:
cv2.imshow('imgThresh', imgThresh)
#all the pixels near boundary will be discarded depending upon the size of kernel. erosion removes white noises
imgThresh = cv2.dilate(imgThresh, kernel_dilate1, iterations=1)
if debug_dilate:
cv2.imshow('dilate-dilate1', imgThresh)
imgThresh = cv2.erode(imgThresh, kernel_erode1, iterations=1)
if debug_erode:
cv2.imshow('dilate-erode1', imgThresh)
imgThresh = cv2.dilate(imgThresh, kernel_dilate2, iterations=1)
if debug_dilate:
cv2.imshow('dilate-dilate2', imgThresh)
imgThresh = cv2.erode(imgThresh, kernel_erode2, iterations=1)
if debug_erode:
cv2.imshow('dilate-erode2', imgThresh)
imgThreshCopy = copy.deepcopy(imgThresh)
# Contours can be explained simply as a curve joining all the continuous points (along the boundary),
# having same color or intensity. The contours are a useful tool for shape analysis and object detection and recognition.
# https://docs.opencv.org/3.1.0/d4/d73/tutorial_py_contours_begin.html
#im2, contours, hierarchy = cv2.findContours(imgThreshCopy, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
imgThreshCopy, contours, hierarchy = cv2.findContours(
imgThreshCopy, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
drawAndShowContours(imgThreshCopy, contours, 'imgContours')
#up here we made all processing image stuff and now we need to work with the info we extrated from the image
#for every thing it's identified on the screen, check if it is a people
for x in contours:
convexHull = cv2.convexHull(x)
blob = Blob(convexHull)
if (blob.isObject()):
currentBlobs.append(blob)
drawAndShowBlobs(imgThresh, currentBlobs, "imgCurrentBlobs")
if atFrame <= 2:
#if it is first iteration there is no comparison, add curBlos to blobs
for curBlob in currentBlobs:
curBlob.id = Blob.getId()
blobs.append(curBlob)
else:
#otherwise check if the curblob is releated to a previous blob and match them
matchCurrentFrameBlobsToExistingBlobs(blobs, currentBlobs)
if debug_all_current_blobs:
for b in blobs:
print b
drawAndShowBlobs(imgThresh, blobs, "imgBlobs")
imgFrame2Copy = copy.deepcopy(imgFrame2)
drawBlobInfoOnImage(blobs, imgFrame2Copy)
#check if the blob crossed the explained
atLeastOneBlobCrossedTheLine = checkIfBlobsCossedTheLine(
blobs, line, peopleCount, seenPeople, case)
#if it has cross draw a colorful line
if atLeastOneBlobCrossedTheLine:
cv2.line(imgFrame2Copy, (line[0].x, line[0].y),
(line[1].x, line[1].y), (255, 0, 255), 2) #yellow line
else:
cv2.line(imgFrame2Copy, (line[0].x, line[0].y),
(line[1].x, line[1].y), (0, 255, 255), 2)
#draw the counter
drawPeopleCounterOnImage(peopleCount, imgFrame2Copy, width, height)
cv2.imshow('imgFrame2Copy', imgFrame2Copy)
# get ready for next iteration
del currentBlobs[:]
imgFrame1 = copy.deepcopy(imgFrame2)
if ((video.get(cv2.CAP_PROP_POS_FRAMES) + 1) <
(video.get(cv2.CAP_PROP_FRAME_COUNT))):
_, imgFrame2 = video.read()
else:
print("end of video")
break
atFrame += 1
#print("frame: " + str(count))
if cv2.waitKey(1) & 0xFF == ord('q'):
break
if debug_mode and cv2.waitKey() & 0xFF == ord('q'):
break
video.release()
cv2.destroyAllWindows()
print("end")
xposv = []
yposv = []
xpos = 0
ypos = 0
if (NFrame):
for x in range(int(fils / divfact)):
for y in range(int(cols / divfact)):
currentColor = frame[divfact * x][divfact * y]
d = np.abs(float(currentColor[0]) - float(Ctrack[0])) + np.abs(
float(currentColor[1]) - float(Ctrack[1])) + np.abs(
float(currentColor[2]) - float(Ctrack[2]))
if (d < Treshold):
if (len(blobs) > 0):
for i in range(len(blobs)):
auxblob = Blob(divfact * x, divfact * y,
blobTreshold)
if (blobs[i].same(auxblob)):
blobs[i].expandB(auxblob)
else:
blobs.append(
Blob(divfact * x, divfact * y,
blobTreshold))
count += 1
else:
blobs.append(
Blob(divfact * x, divfact * y, blobTreshold))
NFrame = False
if (count > 10):
for i in range(len(blobs)):
def addNewBlob(curBlob, blobs):
curBlob.id = Blob.getId()
curBlob.isMatchFoundOrNewBlob = True
blobs.append(curBlob)
def _new_blob(self):
self._last_used_ID += 1
return Blob(3, self._last_used_ID)
# Set up buttons
buttonSpacingCoeff = 1 / 7
buttonYValue = SCREEN_Y * 0.75
buttonXValueMulti = (SCREEN_X - Button.width)
numbBlobsButton = Button("Number of Blobs", (255,255,255), buttonXValueMulti * buttonSpacingCoeff, buttonYValue)
visionButton = Button("Vision Radius", (0,255,0), buttonXValueMulti * buttonSpacingCoeff * 2, buttonYValue)
matingSizeButton = Button("Mating Size", (255,0,0), buttonXValueMulti * buttonSpacingCoeff * 3, buttonYValue)
reachedTargetDistanceButton = Button("Target Dedication", (0,0,255), buttonXValueMulti * buttonSpacingCoeff * 4, buttonYValue)
babySizeButton = Button("Baby Size", (0,255,255), buttonXValueMulti * buttonSpacingCoeff * 5, buttonYValue)
speedButton = Button("Speed", (255,255,0), buttonXValueMulti * buttonSpacingCoeff * 6, buttonYValue)
# Spawn initial Blobs
Blob.screenX = SCREEN_X
Blob.screenY = SCREEN_Y
for i in range(0, 75):
Blob.makeInitialBlob()
# Spawn initial Fruits
for i in range(0, 16):
fruit = Fruit(SCREEN_X, SCREEN_Y)
# Main Game Loop
loops = 0
running = True
while running:
# Event Listeners
for event in pygame.event.get():
# Controls
if event.type == pygame.KEYDOWN:
def __init__(self, x_boundary, y_boundary):
Blob.__init__(self, (0, 0, 255), x_boundary, y_boundary)
logging.info('Blob init with color {}'.format(str(self.color)))
def blob_detector(img, x_i, y_i):
og = img.copy()
img = cv2.Canny(img, 100, 200)
blobs = []
#get connected components of fg
#print ("calculating connected components...")
output = cv2.connectedComponentsWithStats(img, 4, cv2.CV_32S)
# Get the results
# The first cell is the number of labels
#print ("calculating statistics...")
num_labels = output[0]
# The second cell is the label matrix
labels = output[1]
# The third cell is the stat matrix
stats = output[2]
# The fourth cell is the centroid matrix
centroids = output[3]
#print (centroids)
#print(np.where(labels == 2))
roundnesses = []
aspect_ratios = []
formfactors = []
errors = []
rois = []
#imshow_components(labels, og)
#print ("calculating connected components properties...")
# with concurrent.futures.ThreadPoolExecutor(5) as executor:
# future_moles = {executor.submit(compute_stats, num_labels,labels, stats, centroids, img, x_i, y_i, i, og):i for i in range(num_labels)}
# for future in concurrent.futures.as_completed(future_moles):
# found_blobs, roundnesses1, aspect_ratios1, formfactors1, errors1, roi1 = future.result()
# if found_blobs != None:
# blobs.extend(found_blobs)
# roundnesses.append(roundnesses1)
# aspect_ratios.append(aspect_ratios1)
# formfactors.append(formfactors1)
# errors.append(errors1)
# rois.append(roi1)
print("loading model...")
model = load_model(
'/Users/2020shatgiskessell/Desktop/New_Mole_Detector/Mole_Detector_1_3/my_mole_model_2.h5'
)
for i in range(num_labels):
#start = timeit.default_timer()
#get component centroid coordinates
x, y = centroids[i]
#compute statistics
# cv2.CC_STAT_LEFT The leftmost (x) coordinate which is the inclusive start of the bounding box in the horizontal direction.
# cv2.CC_STAT_TOP The topmost (y) coordinate which is the inclusive start of the bounding box in the vertical direction.
# cv2.CC_STAT_WIDTH The horizontal size of the bounding box
# cv2.CC_STAT_HEIGHT The vertical size of the bounding box
# cv2.CC_STAT_AREA The total area (in pixels) of the connected component
width = stats[i, cv2.CC_STAT_WIDTH]
height = stats[i, cv2.CC_STAT_HEIGHT]
if height > 20:
continue
radius = (height + width) / 2
area = stats[i, cv2.CC_STAT_AREA]
#these are the top left x, y coordinates = ONLY TO BE USED FOR GETTING ROI
x_l = stats[i, cv2.CC_STAT_LEFT]
y_l = stats[i, cv2.CC_STAT_TOP]
#compute line
#slope, y_int= compute_lobf(x,y,x_l,y_l)
#stop = timeit.default_timer()
#print('Time to calculate properties: ', stop - start)
#remove everything except for component i to create isolated component matrix
#get connected component roi
roi = og[y_l - 1:y_l + height + 1, x_l - 1:x_l + width + 1]
try:
roi = cv2.resize(roi, (8, 8))
except Exception:
continue
roi = np.expand_dims(roi, axis=2)
roi = np.expand_dims(roi, axis=0)
pred = model.predict(roi)
pred = pred.round()
if int(pred[0][0]) == 1:
print("found blob")
blob = Blob()
blob.radius = radius
blob.x = x + x_i
blob.y = y + y_i
#blob.matrix = component_matrix
blob.roi = roi
blobs.append(blob)
# #----------------MEASURES-------------------------------------------------------------------
#
# #compute more statistics related to roundness
# radius = np.sqrt((area/np.pi))
# formfactor = compute_formfactor (radius, area)
# bounding_box_area_ratio = area/(height*width)
# if height > width:
# roundness = compute_roundness (height, radius, area)
# aspect_ratio = height/width
# else:
# roundness = compute_roundness (width, radius, area)
# aspect_ratio = width/height
#
# #calculates line of best fit and error
# try:
# cord1, cord2, error = compute_lobf(roi, x_l*y_l)
# x1,y1 = cord1
# x2,y2 = cord2
# except TypeError:
# print ("cant calculate line of best fit")
# error = 0
# #COMMENT OUT WHEN COLLECTING ANN DATA!!!!!
# if error < 16:
# continue
return blobs
