def outputs():
global numFrames, time
stream = io.BytesIO()
for i in range(controlFrames):
# This returns the stream for the camera to capture to
e1 = cv2.getTickCount()
yield stream
stream.seek(0)
data = np.fromstring(stream.getvalue(), dtype=np.uint8)
open_cv_image = cv2.imdecode(data, 1)
if filtering:
filt.image = open_cv_image
f, i, c, h = filt.rgbGet(cv2.CHAIN_APPROX_SIMPLE, Constants.VIDEOS_RGB_FILTER_CONSTANTS_1)
output = filt.run(filt.image)
try:
numFrames += 1
# queue.put(open_cv_image)
except:
print "FULL QUEUE"
stream.seek(0)
stream.truncate()
e2 = cv2.getTickCount()
time += (e2-e1)/cv2.getTickFrequency()
def _calculate_metric(self, x, y, transformer, print_row):
tickFrequency = cv2.getTickFrequency()
var_x = numpy.var(x, axis=0)
var_y = numpy.var(y, axis=0)
loss_var_x = numpy.mean(var_x)
loss_var_y = numpy.mean(var_y)
loss = calculate_reconstruction_error(x, y)
matches = match_error(x, y)
start_tick = cv2.getTickCount()
if self.top == 0:
correlation = (calculate_mardia(x, y, 0) / x.shape[1]) * 100
else:
correlation = (calculate_mardia(x, y, self.top) / self.top) * 100
current_time = cv2.getTickCount()
OutputLog().write('calculated correlation, time: {0}'.format(((current_time - start_tick) / tickFrequency)),
'debug')
print_row.append(correlation)
print_row.append(loss)
print_row.append(loss_var_x)
print_row.append(loss_var_y)
print_row.append(matches)
return correlation
def train_data(self, train, target, epochs = 1, log=False):
result = numpy.zeros(epochs)
len_all_inputs = len(train)
for epoch in range(0, epochs):
if log:
print "Executando Epoca %d..." %(epoch+1)
e1 = cv2.getTickCount()
for i, inputs in enumerate(train):
correct = target[i]
output, correct = self.__train(inputs, correct, log)
result[epoch] = result[epoch] + numpy.mean((correct - output)**2)/len_all_inputs
self.__update()
if log:
e2 = getTickCount()
time = (e2 - e1)/cv2.getTickFrequency()
print "Epoca %d Resultado = %f, Tempo de Execucao %f segundos" %(epoch+1,result[epoch], time)
return result
def detect_and_draw( img):
"""
draw a box with opencv on the image around the detected faces and display the output
"""
from random import randint
start_time = cv2.getTickCount()
dog_classifier=DogClassifier()
print dog_classifier
dogs,cats = detectByMuitScaleSlideWindows(img,windowSize=(50,50),classifier=dog_classifier)
end_time = cv2.getTickCount()
logging.info("time cost:%gms",(end_time-start_time)/cv2.getTickFrequency()*1000.)
#dogs = detectObject(img)
print "found dogs:",len(dogs)
max_rect=(0,0,0,0)
for rect in dogs:
x,y,w,h=rect
if w*h>max_rect[2]*max_rect[3]:
max_rect=rect
x,y,w,h=max_rect
cv2.rectangle(img, (x,y), (x+w,y+h), (0,255,255), 1)
print "found cats:",len(cats)
max_rect=(0,0,0,0)
for rect in cats:
x,y,w,h=rect
if w*h>max_rect[2]*max_rect[3]:
max_rect=rect
x,y,w,h=max_rect
cv2.rectangle(img, (x,y), (x+w,y+h), (255,0,0), 1)
cv2.imshow("result", img)
cv2.waitKey(-1)
def bitwise():
img1 = cv2.imread('messi.jpg')
img2 = cv2.imread('python.jpg')
e1 = cv2.getTickCount()
# I want to put logo on top-left corner, So I create a ROI
rows,cols,channels = img2.shape
roi = img1[0:rows, 0:cols]
# Now create a mask of logo and create its inverse mask also
img2gray = cv2.cvtColor(img2,cv2.COLOR_BGR2GRAY)
# cv2.threshold(src, thresh, maxval, type[, dst]) → retval, dst
ret, mask = cv2.threshold(img2gray, 230, 255, cv2.THRESH_BINARY_INV) #Applies a fixed-level threshold to each array element.
mask_inv = cv2.bitwise_not(mask)
# Now black-out the area of logo in ROI
img1_bg = cv2.bitwise_and(roi,roi,mask = mask_inv)
# Take only region of logo from logo image.
img2_fg = cv2.bitwise_and(img2,img2,mask = mask)
# Put logo in ROI and modify the main image
dst = cv2.add(img1_bg,img2_fg)
img1[0:rows, 0:cols] = dst
e2 =cv2.getTickCount()
t = (e2- e1)/cv2.getTickFrequency()
print t
cv2.imshow('res',img1)
cv2.waitKey(0)
cv2.destroyAllWindows()
def get_state(self, motion_detect):
if self.state == 'start_dreaming':
self.__dreaming_start()
elif self.state == 'fade_dream_to_frame':
self.state = 'fading'
elif self.state == 'fading':
self.__fading()
elif motion_detect > self.motion_threshold:
self.__on_motion_above_threshold()
elif cv2.getTickCount() - self.low_motion_last_time > self.LOW_MOTION_TIMEOUT:
self.state = 'waiting'
else:
if motion_detect > self.LOW_THRESHOLD:
self.low_motion_last_time = cv2.getTickCount()
if self.motion:
self.start_time = cv2.getTickCount()
self.motion = False
else:
if self.state == 'dreaming':
how_long = cv2.getTickCount() - self.dream_start
if how_long > self.DREAM_OVER:
self.state = 'fade_dream_to_frame'
elif self.state == 'show_frames':
how_long = cv2.getTickCount() - self.start_time
if how_long > 7000000000:
self.state = 'start_dreaming'
return self.state
def run(self):
"""Run the main loop."""
self._windowManager.createWindow()
while self._windowManager.isWindowCreated:
self._captureManager.enterFrame()
frame = self._captureManager.frame
if frame is not None:
t = cv2.getTickCount()
self._faceTracker.update(frame)
faces = self._faceTracker.faces
t = cv2.getTickCount() - t
print("time taken for detection = %gms" % (t/(cv2.getTickFrequency())*1000.))
# uncomment this line for swapping faces
#rects.swapRects(frame, frame, [face.faceRect for face in faces])
#filters.strokeEdges(frame, frame)
#self._curveFilter.apply(frame, frame)
if self._shouldDrawDebugRects:
self._faceTracker.drawDebugRects(frame)
self._faceTracker.drawLinesFromCenter(frame)
self._captureManager.exitFrame()
self._windowManager.processEvents()
def doTask(self, frame):
time_now = cv2.getTickCount()
time_end = (time_now - self.last_time) / cv2.getTickFrequency()
# Record time value in a rolling sum where the oldest falls off
# This way we can smooth out the FPS value
oldest_val = self.time_records[self.time_rec_idx]
if oldest_val > 0:
self.time_total -= oldest_val
self.time_total += time_end
self.time_records[self.time_rec_idx] = time_end
self.time_rec_idx = (self.time_rec_idx + 1) % NUM_TIME_RECORDS
# Write FPS to frame
#cv2.putText(frame, '%2.2f FPS' % (1 / time_end), (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 1, (255,255,255))
cv2.putText(frame, '%2.2f FPS' % (NUM_TIME_RECORDS / self.time_total), (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 1, (255,255,255))
cv2.imshow("Video", frame)
self.last_time = cv2.getTickCount()
if cv2.waitKey(1) & 0xFF == ord('q'):
return False
else:
return True
def take_snapshot(delay=2):
cap = cv2.VideoCapture(0)
if not cap.isOpened():
print "Cannot open camera!"
return
# Set video to 320x240
cap.set(3, 320)
cap.set(4, 240)
take_picture = False;
t0, filenum = 0, 1
while True:
val, frame = cap.read()
cv2.imshow("video", frame)
key = cv2.waitKey(30)
if key == ord(' '):
t0 = cv2.getTickCount()
take_picture = True
elif key == ord('q'):
break
if take_picture and ((cv2.getTickCount()-t0) / cv2.getTickFrequency()) > delay:
cv2.imwrite(str(filenum) + ".jpg", frame)
filenum += 1
take_picture = False
cap.release()
def capture_camera(mirror=True, size=None):
"""Capture video from camera"""
cap = cv2.VideoCapture(0)
freq = 1000/cv2.getTickFrequency()
start_time = cv2.getTickCount()
oldcnt = 0
cnt = 0
while True:
now_time = cv2.getTickCount()
diff_time = (now_time - start_time)*freq
if diff_time > 1000:
start_time = now_time
fps = cnt - oldcnt
oldcnt = cnt
print fps
ret, frame = cap.read()
if mirror is True:
frame = frame[:,::-1]
if size is not None and len(size) == 2:
frame = cv2.resize(frame, size)
cv2.imshow('camera capture', frame)
k = cv2.waitKey(1)
if k == 27:
break
cnt += 1
cap.release()
cv2.destroyAllWindows()
def compute_outputs(self, test_set_x, test_set_y, hyperparameters):
hidden_output_model = self._build_hidden_model()
recon_model = self._build_reconstruction_model()
hidden_values_x = []
hidden_values_y = []
number_of_batches = int(math.ceil(float(test_set_x.shape[0]) / hyperparameters.batch_size))
outputs_hidden = None
outputs_recon = None
for i in range(number_of_batches):
batch_x = test_set_x[
i * hyperparameters.batch_size: min((i + 1) * hyperparameters.batch_size, test_set_x.shape[0]), :]
batch_y = test_set_y[
i * hyperparameters.batch_size: min((i + 1) * hyperparameters.batch_size, test_set_y.shape[0]), :]
self._correlation_optimizer.var_x.set_value(batch_x)
self._correlation_optimizer.var_y.set_value(batch_y)
start_tick = cv2.getTickCount()
outputs_hidden_batch = hidden_output_model()
output_recon_batch = recon_model()
tickFrequency = cv2.getTickFrequency()
current_time = cv2.getTickCount()
OutputLog().write('batch {0}/{1} ended, time: {2}'.format(i,
number_of_batches,
((current_time - start_tick) / tickFrequency)),
'debug')
if i == 0:
outputs_recon = output_recon_batch
else:
for idx in range(len(outputs_recon)):
outputs_recon[idx] = numpy.concatenate((outputs_recon[idx], output_recon_batch[idx]), axis=0)
if i == 0:
outputs_hidden = outputs_hidden_batch
else:
for idx in range(len(outputs_hidden)):
outputs_hidden[idx] = numpy.concatenate((outputs_hidden[idx], outputs_hidden_batch[idx]), axis=0)
for i in xrange(len(outputs_hidden) / 2):
hidden_values_x.append(outputs_hidden[2 * i])
hidden_values_y.append(outputs_hidden[2 * i + 1])
# for i in xrange(len(outputs_recon) / 2):
# hidden_values_x.append(outputs_recon[2 * i])
# hidden_values_y.append(outputs_recon[2 * i + 1])
# if hidden_values_x[0].shape[1] == hidden_values_y[-1].shape[1] and len(hidden_values_x) > 1:
# hidden_values_x.append(hidden_values_x[-1])
# hidden_values_y.append(hidden_values_y[0])
return [hidden_values_x, hidden_values_y]
def serial_continuous(self, verbose=False):
# assuming only one processor doing work
# so kill others if they exist (they shouldn't though)
if self.rank != 0:
return
self.log("Serial execution, continuous, action={0}".format(self.action))
start_time = getTickCount()
clip_num = 0
while True:
if verbose:
print "Clip: {0}".format(clip_num)
clip_num += 1
in_frames = self.vid.read_next_clip()
self.vid.write_frame_block(self.do_action(in_frames))
# end on last frame block
if len(in_frames) < self.vid.clip_size:
break
self.vid.destroy()
end_time = getTickCount()
comp_time = (end_time - start_time) / getTickFrequency()
self.log("Computation time: {0} s".format(comp_time))
def WeinerFilter():
e1 = cv2.getTickCount()
astro = color.rgb2gray(data.imread(imagePath))
psf = np.ones((5, 5)) / 25
astro = conv2(astro, psf, 'same')
astro += 0.1 * astro.std() * np.random.standard_normal(astro.shape)
deconvolved, _ = restoration.unsupervised_wiener(astro, psf)
fig, ax = plt.subplots(nrows=1, ncols=2, figsize=(8, 5), sharex=True, sharey=True, subplot_kw={'adjustable':'box-forced'})
plt.gray()
ax[0].imshow(astro, vmin=deconvolved.min(), vmax=deconvolved.max())
ax[0].axis('off')
ax[0].set_title('Original Picture')
ax[1].imshow(deconvolved)
ax[1].axis('off')
ax[1].set_title('Self tuned restoration')
fig.subplots_adjust(wspace=0.02, hspace=0.2,
top=0.9, bottom=0.05, left=0, right=1)
e2 = cv2.getTickCount()
time = (e2 - e1)/ cv2.getTickFrequency() + (numberThreads) + numberThreadsPerCore + numberCores
msgbox(msg= str(time) +" seconds", title="Execution Time", ok_button="OK")
plt.show()
def detect_by_slide_windows( img):
"""
draw a box with opencv on the image around the detected faces and display the output
"""
from random import randint
start_time = cv2.getTickCount()
dogcat_classifier=MultiClassifier()
print dogcat_classifier
img=cv2.resize(img,(100,100))
dogs,cats = check_multiLocation_multiScale_windows(img,windowSize=(50,50),classifier=dogcat_classifier)
end_time = cv2.getTickCount()
logging.info("time cost:%gms",(end_time-start_time)/cv2.getTickFrequency()*1000.)
print "found dogs:",len(dogs)
max_rect=(0,0,0,0)
sum_rect=numpy.asarray((0,0,0,0),dtype=numpy.float)
for rect in dogs:
x,y,w,h=rect
sum_rect +=rect
if w*h>max_rect[2]*max_rect[3]:
max_rect=rect
#cv2.rectangle(img, (x,y), (x+w,y+h), (0,255,255), 1)
#x,y,w,h=max_rect
x,y,w,h=numpy.asarray(sum_rect/len(dogs),dtype=numpy.uint)
cv2.rectangle(img, (x,y), (x+w,y+h), (0,255,255), 1)
print "found cats:",len(cats)
max_rect=(0,0,0,0)
for rect in cats:
x,y,w,h=rect
if w*h>max_rect[2]*max_rect[3]:
max_rect=rect
x,y,w,h=max_rect
cv2.rectangle(img, (x,y), (x+w,y+h), (255,0,0), 1)
return img
def Producer(): global timeCapture, numFrames, time try: print "STARTED TO PRODUCE" tot1 = cv2.getTickCount() for frame in camera.capture_continuous(capture, format='bgr', use_video_port=True): # e1 = cv2.getTickCount() # QUEUE FULL EXCEPTION try: queue.put(frame.array) # print "PUT AN IMAGE" except: print "TOO BIG!" # e2 = cv2.getTickCount() # timeCapture += (e2-e1) / cv2.getTickFrequency() capture.truncate(0) if controlFrames != 0 and numFrames >= controlFrames: print "STOPPED PRODUCING" break else: # print numFrames numFrames += 1 except KeyboardInterrupt: # IT HAS ENDED pass tot2 = cv2.getTickCount() total = (tot2 - tot1) / cv2.getTickFrequency() cv2.destroyAllWindows() camera.close() time = 1.0 # CAUSE TIME IS BROKEN RIGHT NOW timeCapture = 1.0 # CAUSE TIME IS COMMENTED OUT ABOVE
def action(self):
''' invoking entry function '''
cap = cv2.VideoCapture(self.video_index)
if not cap.isOpened():
print('ERROR: cannot open video capture device...')
cap.release()
return
self.init_window()
cap.set(cv2.CAP_PROP_FRAME_WIDTH, self.width)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, self.height)
print("press 'q' to quit")
while True:
# Capture frame-by-frame
ret, frame = cap.read()
if not ret:
break
if self.has_skin:
self.skin_mask(frame)
if self.win_count == 0:
#rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
cv2.imshow('frame', frame)
else:
# Our operations on the frame come here
if self.has_gray:
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
cv2.imshow('gray', gray)
if self.has_rgb:
#rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
rgb = frame
cv2.imshow('rgb', rgb)
if self.has_test:
ifrm = frame.copy()
e0 = cv2.getTickCount()
cvimg = hough_lines(ifrm)
e1 = cv2.getTickCount()
elapsed = (e1 - e0) / cv2.getTickFrequency()
show_fps(cvimg, elapsed)
#blue = self.split_blue(ifrm)
cv2.imshow('test', cvimg)
if self.has_hsv:
hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
cv2.imshow('hsv', hsv)
key = cv2.waitKey(1)
if key & 0xFF == ord('q') or key == 27:
break
# When everything done, release the capture
cap.release()
cv2.destroyAllWindows()
def main():
""" Program entry point.
Handles high-level interfacing of video and scene detection / output.
"""
# Get command line arguments directly from the CLI parser defined above.
args = get_cli_parser().parse_args()
# Attempt to open the passed video file as an OpenCV VideoCapture object.
cap = cv2.VideoCapture()
cap.open(args.input.name)
if not cap.isOpened():
print 'FATAL ERROR - could not open video %s.' % args.input.name
print 'cap.isOpened() is not True after calling cap.open(..)'
return
else:
print 'Parsing video %s...' % args.input.name
# Print video parameters (resolution, FPS, etc...)
video_width = cap.get(cv2.CAP_PROP_FRAME_WIDTH)
video_height = cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
video_fps = cap.get(cv2.CAP_PROP_FPS)
print 'Video Resolution / Framerate: %d x %d / %2.3f FPS' % (
video_width, video_height, video_fps )
start_time = cv2.getTickCount() # Record the time we started processing.
if (args.statsfile):
# Only generate statistics, to help setting further parameters.
generate_video_stats(cap, args.statsfile)
else:
# Perform threshold analysis on video, get list of fades in/out.
fade_list = analyze_video_threshold( cap,
args.threshold, args.minpercent, args.blocksize )
# Get # of frames based on position of last frame we read.
frame_count = cap.get(cv2.CAP_PROP_POS_FRAMES)
# Compute & display number of frames, runtime, and average framerate.
total_runtime = float(cv2.getTickCount() - start_time) / cv2.getTickFrequency()
avg_framerate = float(frame_count) / total_runtime
print 'Read %d frames in %4.2f seconds (avg. %4.1f FPS).' % (
frame_count, total_runtime, avg_framerate )
# Ensure we actually detected anything from the video file.
if not len(fade_list) > 0:
print 'Error - no fades detected in video!'
else:
# Generate list of scenes from fades, writing to CSV output if specified.
scene_list = generate_scene_list(cap, fade_list, args.output)
print 'Detected %d scenes in video.' % len(scene_list)
# Cleanup (release all memory and close file handles).
cap.release()
if (args.output): args.output.close()
if (args.statsfile): args.statsfile.close()
print ''
def test_medianBlur(img):
"""docstring for test_medianBlur"""
e1 = cv2.getTickCount()
for i in xrange(5, 49, 2):
img = cv2.medianBlur(img, i)
e2 = cv2.getTickCount()
t = (e2 - e1) / cv2.getTickFrequency()
print t
def send(self, obj, dest, tag):
start = getTickCount()
# MPI Send
self.comm.send(obj, dest=dest, tag=tag)
end = getTickCount()
self.comm_times += (end - start) / getTickFrequency()
def __get_next_frame(self):
delta_t = 0.0
while delta_t < 30000000.0:
t = cv2.getTickCount()
self.video_capture.grab()
delta_t = cv2.getTickCount() - t
rt, frm = self.video_capture.retrieve()
self.last_frame = frm
return frm
def mclass(cls1,cls2,cls3):
imax,jmax=cls1.shape
cls=np.zeros(imax*jmax).reshape(imax,jmax)
t=cv2.getTickCount()
for i in range(imax):
for j in range(jmax):
cls[i,j]=cls1[i,j]+cls2[i,j]*30+cls3[i,j]*900
print (cv2.getTickCount()-t)/cv2.getTickFrequency()
return np.uint16(cls)
def TemporalAlignment(captureQ, captureR):
global harlocsQ, harlocsR;
# if compute_harris == 1:...
totalT1 = float(cv2.getTickCount());
#if config.PREPROCESS_REFERENCE_VIDEO_ONLY == True: # This will give error in multiscale_quad_retrieval since we need to load the saved Harris features anyhow
if True:
# We compute and Store in files the multi-scale Harris features of the reference video
"""
harlocsR = ComputeHarlocs(captureR, config.counterRStep, \
folderName="/harlocs_ref", fileNamePrefix="harloc");
"""
harlocsR = ComputeHarlocs(captureR, config.counterRStep, \
folderName=config.HARRIS_REFERENCE_FOLDER_NAME, \
fileNamePrefix=config.HARRIS_FILENAME_PREFIX, \
fileNameExtension=config.HARRIS_FILENAME_EXTENSION,
indexVideo=1);
if common.MY_DEBUG_STDOUT:
common.DebugPrint("TemporalAlignment(): len(harlocsR) = %s" % str(len(harlocsR)));
sumNbytes = 0;
for hr in harlocsR:
sumNbytes += hr.nbytes;
common.DebugPrint("TemporalAlignment(): harlocsR.nbytes = %s" % str(sumNbytes));
#common.DebugPrint("TemporalAlignment(): harlocsR.nbytes = %s" % str(harlocsR.nbytes));
if config.PREPROCESS_REFERENCE_VIDEO_ONLY == False:
# We compute and Store in files the multi-scale Harris features of the query video
"""
Note: if the "harloc" files exist, load directly the features from
the files without computing them again.
"""
harlocsQ = ComputeHarlocs(captureQ, config.counterQStep, \
folderName=config.HARRIS_QUERY_FOLDER_NAME, \
fileNamePrefix=config.HARRIS_FILENAME_PREFIX, \
fileNameExtension=config.HARRIS_FILENAME_EXTENSION,
indexVideo=0);
if common.MY_DEBUG_STDOUT:
common.DebugPrint("TemporalAlignment(): len(harlocsQ) = %s" % \
str(len(harlocsQ)));
sumNbytes = 0;
for hq in harlocsQ:
sumNbytes += hq.nbytes;
common.DebugPrint("TemporalAlignment(): harlocsQ.nbytes = %s" % str(sumNbytes));
#res = QuadTreeDecision(captureQ, captureR);
res = QuadTreeDecision();
else:
res = None
totalT2 = float(cv2.getTickCount());
myTime = (totalT2 - totalT1) / cv2.getTickFrequency();
print("TemporalAlignment() took %.6f [sec]" % (myTime));
return res;
def __on_faces_present(self):
if not self.faces:
if self.state == 'waiting':
self.start_time = cv2.getTickCount()
elif cv2.getTickCount() - self.start_time > self.FACES_DELAY_TIMER:
if self.state != 'dreaming':
self.state = 'start_dreaming'
self.faces = True
self.faces_latest_time = cv2.getTickCount()
def testPicture(background, inputPic, outputFileName = 'testPicture_out.png'):
t1 = cv2.getTickCount()
output = magic(inputPic,background)#, tolerance=[10,20])
t2 = cv2.getTickCount()
print 'testPicture processed in {0} s'.format((t2 - t1)/cv2.getTickFrequency())
cv2.imwrite(outputFileName, output)
def match(self, grey1, grey2):
word = None
#start tracking time for detection
e1 = cv2.getTickCount()
kpA, descA = self.detector.detectAndCompute(grey1, word)
kpB, descB = self.detector.detectAndCompute(grey2, word)
e2 = cv2.getTickCount()
time = (e2 - e1)/ cv2.getTickFrequency()
return self.bf_matching(kpA, descA, kpB, descB, time)
def medianFilter():
e1 = cv2.getTickCount()
source = cv2.imread(imagePath)
final = cv2.medianBlur(source, 9)
e2 = cv2.getTickCount()
time = (e2 - e1)/ cv2.getTickFrequency() + (numberThreads) + numberThreadsPerCore + numberCores
msgbox(msg= str(time) +" seconds", title="Execution Time", ok_button="OK")
cv2.imshow('Original Picture', source) #Show the image
cv2.imshow('Median Filtered Image', final) #Show the image
cv2.waitKey()
cv2.destroyAllWindows()
def update(self, DEBUG=False, ShowPerformance=False):
def distance(pixel):
return 84134.5 - 0.0390032 * pixel - 53553.8 * np.arctan(pixel ** 2)
def drawReferenceLines():
# 90cm
cv2.line(outputImg, (0, 26), (w, 26), (255))
# 60cm
cv2.line(outputImg, (0, 33), (w, 33), (180))
# 30cm
cv2.line(outputImg, (0, 55), (w, 55), (128))
def drawLine(row, column):
cv2.line(outputImg, (column, row), (column, h), (255))
last = cv2.getTickCount()
while True:
image = q.get()
if image is None:
time.sleep(0.1)
continue
grayImg = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
outputImg = cv2.threshold(grayImg, 160, 255, cv2.THRESH_BINARY)[1]
values = []
for columnIndex in range(0, w, 2):
column = outputImg[:, columnIndex]
row = np.nonzero(column)[0]
if not row.size == 0:
values.append((columnIndex, distance(row[0])))
if DEBUG:
drawLine(row[0], columnIndex)
else:
values.append((columnIndex, 300))
self.values = values
if DEBUG:
drawReferenceLines()
cv2.imshow("image", grayImg)
cv2.imshow('outputImg', outputImg)
cv2.waitKey(1)
if ShowPerformance:
timeSpend = cv2.getTickCount() - last
last = cv2.getTickCount()
timeSec = timeSpend / cv2.getTickFrequency()
print(timeSec, 1 / timeSec)
def main():
img = open_image('../text.bmp')
e1 = cv2.getTickCount()
img = perform_changes_task2(img)
e2 = cv2.getTickCount()
time = (e2 - e1) / cv2.getTickFrequency()
show_image(img)
print "time elapsed: ", time
wait_exit()
def main():
img = open_image('../text.bmp')
e1 = cv2.getTickCount()
img, orig = filter_words(img)
e2 = cv2.getTickCount()
time = (e2 - e1) / cv2.getTickFrequency()
show_image(img, orig)
print "time elapsed: ", time
wait_exit()
def main():
img1 = cv2.imread('1.jpg')
e1 = cv2.getTickCount()
t1 = time.time()
for i in range(5, 49, 2):
img1 = cv2.medianBlur(img1, i)
# your code execution
e2 = cv2.getTickCount()
t2 = time.time()
diff_e = (e2 - e1) / cv2.getTickFrequency()
print(diff_e)
print(t2 - t1)
# if video file successfully open then read frame from video
if (cap.isOpened):
ret, frame = cap.read()
# rescale if specified
if (args.rescale != 1.0):
frame = cv2.resize(frame, (0, 0),
fx=args.rescale,
fy=args.rescale)
# start a timer (to see how long processing and display takes)
start_t = cv2.getTickCount()
# get parameters from track bars
s_lower = cv2.getTrackbarPos("s lower", windowName2)
s_upper = cv2.getTrackbarPos("s upper", windowName2)
v_lower = cv2.getTrackbarPos("v lower", windowName2)
v_upper = cv2.getTrackbarPos("v upper", windowName2)
# select region using the mouse and display it
if (len(boxes) > 1) and (boxes[0][1] < boxes[1][1]) and (boxes[0][0] <
boxes[1][0]):
crop = frame[boxes[0][1]:boxes[1][1],
boxes[0][0]:boxes[1][0]].copy()
def detect_objects(image_np, sess, detection_graph, image, depth, point_cloud,
category_index):
# Expand dimensions since the model expects images to have shape: [1, None, None, 3]
image_np_expanded = np.expand_dims(image_np, axis=0)
image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
# Each box represents a part of the image where a particular object was detected.
boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
# Each score represent how level of confidence for each of the objects.
# Score is shown on the result image, together with the class label.
scores = detection_graph.get_tensor_by_name('detection_scores:0')
classes = detection_graph.get_tensor_by_name('detection_classes:0')
num_detections = detection_graph.get_tensor_by_name('num_detections:0')
t1 = cv2.getTickCount()
# Actual detection.
out = sess.run([boxes, scores, classes, num_detections],
feed_dict={image_tensor: image_np_expanded})
(boxes, scores, classes, num_detections) = out
centerx = 0
centery = 0
key = ''
rows = image_np.shape[0]
cols = image_np.shape[1]
num_detections = int(num_detections)
for i in range(num_detections):
classId = (out[2][0][i])
score = float(out[1][0][i])
bbox = [float(v) for v in out[0][0][i]]
if score > 0.3:
x = bbox[1] * cols
y = bbox[0] * rows
right = bbox[3] * cols
bottom = bbox[2] * rows
#cv2.rectangle(img, (int(x), int(y)), (int(right), int(bottom)), (125, 255, 51), thickness=1)
cv2.circle(image_np, (int((x + right) / 2), int((y + bottom) / 2)),
8, (125, 155, 21),
thickness=1)
centerx = int((x + right) / 2)
centery = int((y + bottom) / 2)
# Get and print distance value in mm at the center of the object
# We measure the distance camera - object using Euclidean distance
err, point_cloud_value = point_cloud.get_value(centerx, centery)
distance = math.sqrt(point_cloud_value[0] * point_cloud_value[0] +
point_cloud_value[1] * point_cloud_value[1] +
point_cloud_value[2] * point_cloud_value[2])
if not np.isnan(distance) and not np.isinf(distance):
distance = round(distance)
#print("object {0} Distance to Camera at ({1}, {2}): {3} mm\n".format(i,x, y, distance))
classes = np.squeeze(classes).astype(np.int32)
if classes[i] in category_index.keys():
class_name = category_index[classes[i]]['name']
else:
class_name = 'N/A'
display_str = str(class_name)
cv2.putText(image_np,
"{0}:{1} mm".format(display_str, distance),
(int(centerx), int(centery)),
cv2.FONT_HERSHEY_SIMPLEX,
1, (125, 155, 21),
thickness=2,
lineType=2)
# Increment the loop
else:
print(
"Can't estimate distance at this position, move the camera\n"
)
sys.stdout.flush()
t2 = cv2.getTickCount()
print((t2 - t1) / cv2.getTickFrequency())
return image_np
def main():
model_name = 'siamRPN'
snapshot_path = os.path.join(project_path_, 'experiments/%s/model.pth' % args.tracker_name)
config_path = os.path.join(project_path_, 'experiments/%s/config.yaml' % args.tracker_name)
cfg.merge_from_file(config_path)
# create model
model = ModelBuilder() # a model is a Neural Network.(a torch.nn.Module)
# load model
model = load_pretrain(model, snapshot_path).cuda().eval()
# build tracker
tracker = build_tracker(model) # a tracker is a object consisting of not only a NN and some post-processing
# create dataset
dataset = DatasetFactory.create_dataset(name=args.dataset, dataset_root=dataset_root_, load_img=False)
total_lost = 0
if args.dataset in ['VOT2016', 'VOT2018', 'VOT2019']:
# restart tracking
for v_idx, video in enumerate(dataset):
if args.video != '':
# test one special video
if video.name != args.video:
continue
frame_counter = 0
lost_number = 0
toc = 0
pred_bboxes = []
for idx, (img, gt_bbox) in enumerate(video):
if len(gt_bbox) == 4:
gt_bbox = [gt_bbox[0], gt_bbox[1],
gt_bbox[0], gt_bbox[1]+gt_bbox[3]-1,
gt_bbox[0]+gt_bbox[2]-1, gt_bbox[1]+gt_bbox[3]-1,
gt_bbox[0]+gt_bbox[2]-1, gt_bbox[1]]
tic = cv2.getTickCount()
if idx == frame_counter:
H,W,_ = img.shape
cx, cy, w, h = get_axis_aligned_bbox(np.array(gt_bbox))
gt_bbox_ = [cx-(w-1)/2, cy-(h-1)/2, w, h]
tracker.init(img, gt_bbox_)
pred_bbox = gt_bbox_
pred_bboxes.append(1)
elif idx > frame_counter:
outputs = tracker.track(img)
pred_bbox = outputs['bbox']
overlap = vot_overlap(pred_bbox, gt_bbox, (img.shape[1], img.shape[0]))
if overlap > 0:
# not lost
pred_bboxes.append(pred_bbox)
else:
# lost object
pred_bboxes.append(2)
frame_counter = idx + 5 # skip 5 frames
lost_number += 1
else:
pred_bboxes.append(0)
toc += cv2.getTickCount() - tic
if idx == 0:
cv2.destroyAllWindows()
if args.vis and idx > frame_counter:
cv2.polylines(img, [np.array(gt_bbox, np.int).reshape((-1, 1, 2))],
True, (0, 255, 0), 3)
bbox = list(map(int, pred_bbox))
cv2.rectangle(img, (bbox[0], bbox[1]),
(bbox[0]+bbox[2], bbox[1]+bbox[3]), (0, 255, 255), 3)
cv2.putText(img, str(idx), (40, 40), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 255), 2)
cv2.putText(img, str(lost_number), (40, 80), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)
cv2.imshow(video.name, img)
cv2.waitKey(1)
toc /= cv2.getTickFrequency()
# save results
video_path = os.path.join(save_dir, args.dataset, model_name,
'baseline', video.name)
if not os.path.isdir(video_path):
os.makedirs(video_path)
result_path = os.path.join(video_path, '{}_001.txt'.format(video.name))
with open(result_path, 'w') as f:
for x in pred_bboxes:
if isinstance(x, int):
f.write("{:d}\n".format(x))
else:
f.write(','.join([vot_float2str("%.4f", i) for i in x])+'\n')
print('({:3d}) Video: {:12s} Time: {:4.1f}s Speed: {:3.1f}fps Lost: {:d}'.format(
v_idx+1, video.name, toc, idx / toc, lost_number))
def track(tracker, net, video, args):
start_frame, lost_times, toc = 0, 0, 0
# save result to evaluate
if args.epoch_test:
suffix = args.resume.split('/')[-1]
suffix = suffix.split('.')[0]
tracker_path = os.path.join('result', args.dataset, args.arch + suffix)
else:
tracker_path = os.path.join('result', args.dataset, args.arch)
if not os.path.exists(tracker_path):
os.makedirs(tracker_path)
if 'VOT' in args.dataset:
baseline_path = join(tracker_path, 'baseline')
video_path = join(baseline_path, video['name'])
if not os.path.exists(video_path):
os.makedirs(video_path)
result_path = join(video_path, video['name'] + '_001.txt')
else:
result_path = join(tracker_path, '{:s}.txt'.format(video['name']))
if os.path.exists(result_path):
return 0 # for mult-gputesting
regions = [] # result and states[1 init / 2 lost / 0 skip]
image_files, gt = video['image_files'], video['gt']
for f, image_file in enumerate(image_files):
im = cv2.imread(image_file)
if len(im.shape) == 2:
im = cv2.cvtColor(im, cv2.COLOR_GRAY2BGR)
tic = cv2.getTickCount()
if f == start_frame: # init
cx, cy, w, h = get_axis_aligned_bbox(gt[f])
target_pos = np.array([cx, cy])
target_sz = np.array([w, h])
state = tracker.init(im, target_pos, target_sz,
net) # init tracker
regions.append(1 if 'VOT' in args.dataset else gt[f])
elif f > start_frame: # tracking
state = tracker.track(state, im) # track
location = cxy_wh_2_rect(state['target_pos'], state['target_sz'])
b_overlap = poly_iou(gt[f],
location) if 'VOT' in args.dataset else 1
if b_overlap > 0:
regions.append(location)
else:
regions.append(2)
lost_times += 1
start_frame = f + 5 # skip 5 frames
else: # skip
regions.append(0)
toc += cv2.getTickCount() - tic
toc /= cv2.getTickFrequency()
with open(result_path, "w") as fin:
if 'VOT' in args.dataset:
for x in regions:
if isinstance(x, int):
fin.write("{:d}\n".format(x))
else:
p_bbox = x.copy()
fin.write(','.join([str(i) for i in p_bbox]) + '\n')
else:
for x in regions:
p_bbox = x.copy()
fin.write(','.join([
str(i + 1) if idx == 0 or idx == 1 else str(i)
for idx, i in enumerate(p_bbox)
]) + '\n')
print('Video: {:12s} Time: {:2.1f}s Speed: {:3.1f}fps Lost: {:d}'.format(
video['name'], toc, f / toc, lost_times))
return lost_times
cap = cv2.VideoCapture(1)
# TRACKER INITIALIZATION
success, frame = cap.read()
bbox = cv2.selectROI("Tracking",frame, False)
tracker.init(frame, bbox)
def drawBox(img,bbox):
x, y, w, h = int(bbox[0]), int(bbox[1]), int(bbox[2]), int(bbox[3])
cv2.rectangle(img, (x, y), ((x + w), (y + h)), (255, 0, 255), 3, 3 )
cv2.putText(img, "Tracking", (100, 75), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 255, 0), 2)
while True:
timer = cv2.getTickCount()
success, img = cap.read()
success, bbox = tracker.update(img)
if success:
drawBox(img,bbox)
else:
cv2.putText(img, "Lost", (100, 75), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
cv2.rectangle(img,(15,15),(200,90),(255,0,255),2)
cv2.putText(img, "Fps:", (20, 40), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (255,0,255), 2);
cv2.putText(img, "Status:", (20, 75), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (255, 0, 255), 2);
fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer);
if fps>60: myColor = (20,230,20)
def count(self, frame):
_timer = cv2.getTickCount(
) # set timer to calculate processing frame rate
self.frame = frame
for _id, blob in list(self.blobs.items()):
# update trackers
success, box = blob.tracker.update(self.frame)
if success:
blob.num_consecutive_tracking_failures = 0
blob.update(box)
logger.debug('Vehicle tracker updated.',
extra={
'meta': {
'cat': 'TRACKER_UPDATE',
'vehicle_id': _id,
'bounding_box': blob.bounding_box,
'centroid': blob.centroid,
},
})
else:
blob.num_consecutive_tracking_failures += 1
# count vehicles that have left the frame if no counting line exists
# or those that have passed the counting line if one exists
if (self.counting_line == None and \
(blob.num_consecutive_tracking_failures == self.mctf or blob.num_consecutive_detection_failures == self.mcdf) and \
not blob.counted) \
or \
(self.counting_line != None and \
# don't count a blob if it was first detected at a position past the counting line
# this enforces counting in only one direction
not is_passed_counting_line(blob.position_first_detected, self.counting_line, self.cl_position) and \
is_passed_counting_line(blob.centroid, self.counting_line, self.cl_position) and \
not blob.counted):
blob.counted = True
self.vehicle_count += 1
# count by vehicle type
if blob.type != None:
if blob.type in self.types_counts:
self.types_counts[blob.type] += 1
else:
self.types_counts[blob.type] = 1
logger.info('Vehicle counted.',
extra={
'meta': {
'cat': 'VEHICLE_COUNT',
'id': _id,
'type': blob.type,
'count': self.vehicle_count,
'position_first_detected':
blob.position_first_detected,
'position_counted': blob.centroid,
'counted_at': time.time(),
},
})
if blob.num_consecutive_tracking_failures >= self.mctf:
# delete untracked blobs
del self.blobs[_id]
if self.frame_count >= self.di:
# rerun detection
droi_frame = get_roi_frame(self.frame, self.droi)
_bounding_boxes, _classes, _confidences = get_bounding_boxes(
droi_frame, self.detector)
self.blobs = add_new_blobs(_bounding_boxes, _classes, _confidences,
self.blobs, self.frame, self.tracker,
self.counting_line, self.cl_position,
self.mcdf)
self.blobs = remove_duplicates(self.blobs)
self.frame_count = 0
self.frame_count += 1
self.processing_frame_rate = round(
cv2.getTickFrequency() / (cv2.getTickCount() - _timer), 2)
logger.debug('Processing frame rate updated.',
extra={
'meta': {
'cat': 'PROCESSING_SPEED',
'frame_rate': self.processing_frame_rate
},
})
def run_tracker(tracker, img, gt, video_name, restart=True):
frame_counter = 0
lost_number = 0
toc = 0
pred_bboxes = []
if restart: # VOT2016 and VOT 2018
for idx, (img, gt_bbox) in enumerate(video):
if len(gt_bbox) == 4:
gt_bbox = [
gt_bbox[0], gt_bbox[1], gt_bbox[0],
gt_bbox[1] + gt_bbox[3] - 1, gt_bbox[0] + gt_bbox[2] - 1,
gt_bbox[1] + gt_bbox[3] - 1, gt_bbox[0] + gt_bbox[2] - 1,
gt_bbox[1]
]
tic = cv2.getTickCount()
if idx == frame_counter:
cx, cy, w, h = get_axis_aligned_bbox(np.array(gt_bbox))
gt_bbox_ = [cx - (w - 1) / 2, cy - (h - 1) / 2, w, h]
tracker.init(img, gt_bbox_)
pred_bbox = gt_bbox_
pred_bboxes.append(1)
elif idx > frame_counter:
outputs = tracker.track(img)
pred_bbox = outputs['bbox']
overlap = vot_overlap(pred_bbox, gt_bbox,
(img.shape[1], img.shape[0]))
if overlap > 0:
# not lost
pred_bboxes.append(pred_bbox)
else:
# lost object
pred_bboxes.append(2)
frame_counter = idx + 5 # skip 5 frames
lost_number += 1
else:
pred_bboxes.append(0)
toc += cv2.getTickCount() - tic
toc /= cv2.getTickFrequency()
print(
'Video: {:12s} Time: {:4.1f}s Speed: {:3.1f}fps Lost: {:d}'.format(
video_name, toc, idx / toc, lost_number))
return pred_bboxes
else:
toc = 0
pred_bboxes = []
scores = []
track_times = []
for idx, (img, gt_bbox) in enumerate(video):
tic = cv2.getTickCount()
if idx == 0:
cx, cy, w, h = get_axis_aligned_bbox(np.array(gt_bbox))
gt_bbox_ = [cx - (w - 1) / 2, cy - (h - 1) / 2, w, h]
tracker.init(img, gt_bbox_)
pred_bbox = gt_bbox_
scores.append(None)
pred_bboxes.append(pred_bbox)
else:
outputs = tracker.track(img)
pred_bbox = outputs['bbox']
pred_bboxes.append(pred_bbox)
scores.append(outputs['best_score'])
toc += cv2.getTickCount() - tic
track_times.append(
(cv2.getTickCount() - tic) / cv2.getTickFrequency())
toc /= cv2.getTickFrequency()
print('Video: {:12s} Time: {:5.1f}s Speed: {:3.1f}fps'.format(
video_name, toc, idx / toc))
return pred_bboxes, scores, track_times
import numpy as np
import cv2
from matplotlib import pyplot as plt
#Lecture image en niveau de gris et conversion en float64
img = np.float64(cv2.imread('../Image_Pairs/FlowerGarden2.png', 0))
(h, w) = img.shape
print("Dimension de l'image :", h, "lignes x", w, "colonnes")
#Méthode directe
t1 = cv2.getTickCount()
img2 = cv2.copyMakeBorder(img, 0, 0, 0, 0, cv2.BORDER_REPLICATE)
for y in range(1, h - 1):
for x in range(1, w - 1):
val = 5 * img[y, x] - img[y - 1,
x] - img[y, x - 1] - img[y + 1,
x] - img[y, x + 1]
img2[y, x] = min(max(val, 0), 255)
t2 = cv2.getTickCount()
time = (t2 - t1) / cv2.getTickFrequency()
print("Méthode directe :", time, "s")
plt.subplot(121)
plt.imshow(img2, cmap='gray')
plt.title('Convolution - Méthode Directe')
#Méthode filter2D
t1 = cv2.getTickCount()
kernel = np.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]])
img3 = cv2.filter2D(img, -1, kernel)
def main():
# Create the in-memory stream
stream = io.BytesIO()
# Create a window
cv2.namedWindow(WINDOW_DISPLAY_IMAGE)
# position the window
cv2.moveWindow(WINDOW_DISPLAY_IMAGE, 0, 35)
# Add some controls to the window
cv2.createTrackbar(CONTROL_SCAN_RADIUS, WINDOW_DISPLAY_IMAGE, 5, 50,
onScanRadiusChange)
cv2.setTrackbarPos(CONTROL_SCAN_RADIUS, WINDOW_DISPLAY_IMAGE,
SCAN_RADIUS_REG)
cv2.createTrackbar(CONTROL_NUMBER_OF_CIRCLES, WINDOW_DISPLAY_IMAGE, 0, 7,
onCircleScanChange)
cv2.setTrackbarPos(CONTROL_NUMBER_OF_CIRCLES, WINDOW_DISPLAY_IMAGE,
NUMBER_OF_CIRCLES)
cv2.createTrackbar(CONTROL_LINE_WIDTH, WINDOW_DISPLAY_IMAGE, 0,
RESOLUTION_X, onLineWidthChange)
cv2.setTrackbarPos(CONTROL_LINE_WIDTH, WINDOW_DISPLAY_IMAGE,
SCAN_RADIUS * 2)
returnString = """
Press Esc to end the program
Using camera resolution: {}x{}
Centre point: {}:{}
Scan radius: {}
Number of search itterations: {}
Baseline Width: {}
""".format(RESOLUTION_X, RESOLUTION_Y, SCAN_POS_X, SCAN_HEIGHT,
SCAN_RADIUS_REG, NUMBER_OF_CIRCLES, SCAN_RADIUS * 2)
print(returnString)
# Open connection to camera
with picamera.PiCamera() as camera:
# Set camera resolution
camera.resolution = (RESOLUTION_X, RESOLUTION_Y)
# Start loop
while True:
# Get the tick count so we can keep track of performance
e1 = cv2.getTickCount()
# Capture image from camera
camera.capture(stream, format='jpeg', use_video_port=True)
# Convert image from camera to a numpy array
data = np.fromstring(stream.getvalue(), dtype=np.uint8)
# Decode the numpy array image
image = cv2.imdecode(data, cv2.CV_LOAD_IMAGE_COLOR)
# Empty and return the in-memory stream to beginning
stream.seek(0)
stream.truncate(0)
# Create other images
grey_image = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
display_image = cv2.copyMakeBorder(image, 0, 0, 0, 0,
cv2.BORDER_REPLICATE)
center_point = (SCAN_POS_X, SCAN_HEIGHT)
# San a horizontal line based on the centre point
# We could just use this data to work out how far off centre we are and steer accordingly.
# Get a data array of all the falues along that line
# scan_data is an array containing:
# - pixel value
scan_data = scanLine(grey_image, display_image, center_point,
SCAN_RADIUS)
# The center point we believe the line we are following intersects with our scan line.
point_on_line = findLine(display_image, scan_data, SCAN_POS_X,
SCAN_HEIGHT, SCAN_RADIUS)
# Start scanning the arcs
# This allows us to look ahead further ahead at the line and work out an angle to steer
# From the intersection point of the line, scan in an arc to find the line
# The scan data contains an array
# - pixel value
# - x co-ordinate
# - y co-ordinate
returnVal, scan_data = scanCircle(grey_image, display_image,
point_on_line, SCAN_RADIUS_REG,
-90)
previous_point = point_on_line
# in the same way ads the findLine, go through the data, find the mid point and return the co-ordinates.
last_point = findInCircle(display_image, scan_data)
cv2.line(display_image, (previous_point[0], previous_point[1]),
(last_point[0], last_point[1]), (255, 255, 255), 1)
actual_number_of_circles = 0
for scan_count in range(0, NUMBER_OF_CIRCLES):
returnVal, scan_data = scanCircle(
grey_image, display_image, last_point, SCAN_RADIUS_REG,
lineAngle(previous_point, last_point))
# Only work out the next itteration if our point is within the bounds of the image
if returnVal == True:
actual_number_of_circles += 1
previous_point = last_point
last_point = findInCircle(display_image, scan_data)
cv2.line(display_image,
(previous_point[0], previous_point[1]),
(last_point[0], last_point[1]), (255, 255, 255),
1)
else:
break
# Draw a line from the centre point to the end point where we last found the line we are following
cv2.line(display_image, (center_point[0], center_point[1]),
(last_point[0], last_point[1]), (0, 0, 255), 1)
# Display the image
cv2.imshow(WINDOW_DISPLAY_IMAGE, display_image)
# This is the maximum distance the end point of our search for a line can be from the centre point.
line_scan_length = SCAN_RADIUS_REG * (actual_number_of_circles + 1)
# This is the measured line length from the centre point
line_length_from_center = lineLength(center_point, last_point)
center_y_distance = center_point[1] - last_point[1]
center_x_distance = center_point[0] - last_point[0]
# Stop counting all work is done at this point and calculate how we are doing.
e2 = cv2.getTickCount()
returnString = "fps {} - bearing {} - x:{} y:{} look distance:{} distance from origin:{}".format(
1000 / ((e2 - e1) / cv2.getTickFrequency() * 1000),
lineAngle(center_point, last_point) * -1 - 90,
center_x_distance, center_y_distance, line_scan_length,
line_length_from_center)
print(returnString)
# Wait for ESC to end program
c = cv2.waitKey(7) % 0x100
if c == 27:
break
print "Closing program"
cv2.destroyAllWindows()
print "All windows should be closed"
return
def clock():
return cv2.getTickCount() / cv2.getTickFrequency()
def _infer(self, prep_img):
t0 = cv2.getTickCount()
output = self._exec_model.infer(inputs={self._input_layer_name: prep_img})
self.infer_time = (cv2.getTickCount() - t0) / cv2.getTickFrequency()
return output
def calibrate_stereo_cameras(res_x=img_width, res_y=img_height):
# We need a lot of variables to calibrate the stereo camera
"""
Based on code from:
https://gist.github.com/aarmea/629e59ac7b640a60340145809b1c9013
"""
processing_time01 = cv2.getTickCount()
objectPoints = None
rightImagePoints = None
rightCameraMatrix = None
rightDistortionCoefficients = None
leftImagePoints = None
leftCameraMatrix = None
leftDistortionCoefficients = None
rotationMatrix = None
translationVector = None
imageSize = (res_x, res_y)
TERMINATION_CRITERIA = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER,
30, 0.01)
OPTIMIZE_ALPHA = 0.25
try:
npz_file = np.load(
'./calibration_data/{}p/stereo_camera_calibration.npz'.format(
res_y))
except:
pass
for cam_num in [0, 1]:
right_or_left = ["_right" if cam_num == 1 else "_left"][0]
try:
print('./calibration_data/{}p/camera_calibration{}.npz'.format(
res_y, right_or_left))
npz_file = np.load(
'./calibration_data/{}p/camera_calibration{}.npz'.format(
res_y, right_or_left))
list_of_vars = [
'map1', 'map2', 'objpoints', 'imgpoints', 'camera_matrix',
'distortion_coeff'
]
print(sorted(npz_file.files))
if sorted(list_of_vars) == sorted(npz_file.files):
print("Camera calibration data has been found in cache.")
map1 = npz_file['map1']
map2 = npz_file['map2']
objectPoints = npz_file['objpoints']
if right_or_left == "_right":
rightImagePoints = npz_file['imgpoints']
rightCameraMatrix = npz_file['camera_matrix']
rightDistortionCoefficients = npz_file['distortion_coeff']
if right_or_left == "_left":
leftImagePoints = npz_file['imgpoints']
leftCameraMatrix = npz_file['camera_matrix']
leftDistortionCoefficients = npz_file['distortion_coeff']
else:
print("Camera data file found but data corrupted.")
except:
#If the file doesn't exist
print("Camera calibration data not found in cache.")
return False
print("Calibrating cameras together...")
leftImagePoints = np.asarray(leftImagePoints, dtype=np.float64)
rightImagePoints = np.asarray(rightImagePoints, dtype=np.float64)
# Stereo calibration
(RMS, _, _, _, _, rotationMatrix,
translationVector) = cv2.fisheye.stereoCalibrate(
objectPoints, leftImagePoints, rightImagePoints, leftCameraMatrix,
leftDistortionCoefficients, rightCameraMatrix,
rightDistortionCoefficients, imageSize, None, None,
cv2.CALIB_FIX_INTRINSIC, TERMINATION_CRITERIA)
# Print RMS result (for calibration quality estimation)
print("<><><><><><><><><><><><><><><><><><><><>")
print("<><> RMS is ", RMS, " <><>")
print("<><><><><><><><><><><><><><><><><><><><>")
print("Rectifying cameras...")
R1 = np.zeros([3, 3])
R2 = np.zeros([3, 3])
P1 = np.zeros([3, 4])
P2 = np.zeros([3, 4])
Q = np.zeros([4, 4])
# Rectify calibration results
(leftRectification, rightRectification, leftProjection, rightProjection,
dispartityToDepthMap) = cv2.fisheye.stereoRectify(
leftCameraMatrix, leftDistortionCoefficients, rightCameraMatrix,
rightDistortionCoefficients, imageSize, rotationMatrix,
translationVector, 0, R2, P1, P2, Q, cv2.CALIB_ZERO_DISPARITY, (0, 0),
0, 0)
# Saving calibration results for the future use
print("Saving calibration...")
leftMapX, leftMapY = cv2.fisheye.initUndistortRectifyMap(
leftCameraMatrix, leftDistortionCoefficients, leftRectification,
leftProjection, imageSize, cv2.CV_16SC2)
rightMapX, rightMapY = cv2.fisheye.initUndistortRectifyMap(
rightCameraMatrix, rightDistortionCoefficients, rightRectification,
rightProjection, imageSize, cv2.CV_16SC2)
np.savez_compressed(
'./calibration_data/{}p/stereo_camera_calibration.npz'.format(res_y),
imageSize=imageSize,
leftMapX=leftMapX,
leftMapY=leftMapY,
rightMapX=rightMapX,
rightMapY=rightMapY,
dispartityToDepthMap=dispartityToDepthMap)
return True
def run(self):
# Capture, process, display loop
while True:
# Read a new frame
ok, self.frame = self.video.read()
w, h, ch = self.frame.shape
self.frame = cv2.resize(self.frame, (int(h/2), int(w/2)), interpolation=cv2.INTER_NEAREST)
display = self.frame.copy() # Frame we'll do all the graphical drawing to
data_display = np.zeros_like(display, dtype=np.uint8) # Black screen to display data
if not ok:
break
# Start FPS timer
timer = cv2.getTickCount()
if self.bg is None:
cv2.putText(display,
"Press 'r' to reset the foreground extraction.",
self.positions['hand_pose'],
cv2.FONT_HERSHEY_SIMPLEX,
0.75, (0, 127, 64), 2)
cv2.imshow("display", display)
k = cv2.waitKey(1) & 0xff
if k == 27: break # ESC pressed
elif k == 114 or k == 108:
# r pressed
self.bg = self.frame.copy()
self.hand_bbox = (116, 116, 170, 170)
self.is_tracking = False
else:
# Extract the foreground
foreground, mask = self.extract_foreground()
foreground_display = foreground.copy()
# Get hand from mask using the bounding box
hand_crop = mask[int(self.hand_bbox[1]):int(self.hand_bbox[1]+self.hand_bbox[3]),
int(self.hand_bbox[0]):int(self.hand_bbox[0]+self.hand_bbox[2])]
# Update tracker
if self.is_tracking:
tracking, self.hand_bbox = self.tracker.update(foreground)
try:
# Resize cropped hand and make prediction on gesture
hand_crop_resized = np.expand_dims(cv2.resize(hand_crop, (54, 54)), axis=0).reshape((1, 54, 54, 1))
prediction = self.recognizer.predict(hand_crop_resized)
predi = prediction[0].argmax() # Get the index of the greatest confidence
gesture = self.CLASSES[predi]
for i, pred in enumerate(prediction[0]):
# Draw confidence bar for each gesture
barx = self.positions['hand_pose'][0]
bary = 60 + i*60
bar_height = 20
bar_length = int(400 * pred) + barx # calculate length of confidence bar
# Make the most confidence prediction green
if i == predi:
colour = (0, 255, 0)
else:
colour = (0, 0, 255)
cv2.putText(data_display,
"{}: {}".format(self.CLASSES[i], pred),
(self.positions['hand_pose'][0], 30 + i*60),
cv2.FONT_HERSHEY_SIMPLEX,
0.75, (255, 255, 255), 2)
cv2.rectangle(data_display,
(barx, bary),
(bar_length,
bary - bar_height),
colour,
-1, 1)
cv2.putText(display,
"hand pose: {}".format(gesture),
self.positions['hand_pose'],
cv2.FONT_HERSHEY_SIMPLEX,
0.75, (0, 0, 255), 2)
cv2.putText(foreground_display,
"hand pose: {}".format(gesture),
self.positions['hand_pose'],
cv2.FONT_HERSHEY_SIMPLEX,
0.75, (0, 0, 255), 2)
except Exception as ex:
cv2.putText(display,
"hand pose: error",
self.positions['hand_pose'],
cv2.FONT_HERSHEY_SIMPLEX,
0.75, (0, 0, 255), 2)
cv2.putText(foreground_display,
"hand pose: error",
self.positions['hand_pose'],
cv2.FONT_HERSHEY_SIMPLEX,
0.75, (0, 0, 255), 2)
# Draw bounding box
p1 = (int(self.hand_bbox[0]), int(self.hand_bbox[1]))
p2 = (int(self.hand_bbox[0] + self.hand_bbox[2]), int(self.hand_bbox[1] + self.hand_bbox[3]))
cv2.rectangle(foreground_display, p1, p2, (255, 0, 0), 2, 1)
cv2.rectangle(display, p1, p2, (255, 0, 0), 2, 1)
# Calculate difference in hand position
hand_pos = ((p1[0] + p2[0])//2, (p1[1] + p2[1])//2)
mouse_change = ((p1[0] + p2[0])//2 - self.positions['null_pos'][0], self.positions['null_pos'][0] - (p1[1] + p2[1])//2)
# Draw hand moved difference
cv2.circle(display, self.positions['null_pos'], 5, (0,0,255), -1)
cv2.circle(display, hand_pos, 5, (0,255,0), -1)
cv2.line(display, self.positions['null_pos'], hand_pos, (255,0,0),5)
# Calculate Frames per second (FPS)
fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer)
# Display FPS on frame
cv2.putText(foreground_display, "FPS : " + str(int(fps)), self.positions['fps'], cv2.FONT_HERSHEY_SIMPLEX, 0.65, (50, 170, 50), 2)
cv2.putText(display, "FPS : " + str(int(fps)), self.positions['fps'], cv2.FONT_HERSHEY_SIMPLEX, 0.65, (50, 170, 50), 2)
# Display pause command text
cv2.putText(foreground_display,
"hold 'r' to recalibrate until the screen is black",
(15, 400),
cv2.FONT_HERSHEY_SIMPLEX,
0.65, (0, 0, 255), 2)
cv2.putText(foreground_display,
"to recalibrate",
(15, 420),
cv2.FONT_HERSHEY_SIMPLEX,
0.65, (0, 0, 255), 2)
cv2.putText(foreground_display,
"press 'p' to return to paused state",
(15, 450),
cv2.FONT_HERSHEY_SIMPLEX,
0.65, (0, 0, 255), 2)
# Display foreground_display
cv2.imshow("foreground_display", foreground_display)
cv2.imshow("display", display)
cv2.imshow("hand_crop", hand_crop)
# Display result
cv2.imshow("data", data_display)
k = cv2.waitKey(1) & 0xff
if k == 27: break # ESC pressed
elif k == 114 or k == 108: # r pressed
self.bg = self.frame.copy()
self.hand_bbox = (116, 116, 170, 170)
self.is_tracking = False
elif k == 116: # t pressed
# Initialize tracker with first frame and bounding box
self.is_tracking = True
self.tracker = cv2.TrackerKCF_create()
self.tracker.init(self.frame, self.hand_bbox)
elif k == 112: # p pressed
# Reset to paused state
self.is_tracking = False
self.bg = None
cv2.destroyAllWindows()
elif k != 255: print(k)
desc = stack_array([img_data.bow_descriptors for img_data in imgs_data])
desc = np.float32(desc)
# clustering
criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER,
params.BOW_clustering_iterations, 0.01)
flags = cv2.KMEANS_PP_CENTERS
compactness, labels, dictionary = cv2.kmeans(desc, dictionary_size, None,
criteria, 1, flags)
np.save(params.BOW_DICT_PATH, dictionary)
return dictionary
program_start = cv2.getTickCount()
# get training set
print("Loading images...")
start = cv2.getTickCount()
# must specify data path names in same order as class names
paths = [params.DATA_training_path_neg, params.DATA_training_path_pos]
# get class names
class_names = [
get_class_name(class_number)
for class_number in range(len(params.DATA_CLASS_NAMES))
]
# again, must specify in same order as class names
sampling_sizes = [
params.DATA_training_sample_count_neg,
params.DATA_training_sample_count_pos
'''
3. Performance-measurement and -improvement
'''
import cv2
import numpy as np
import matplotlib.pyplot as plt
# The modules time and profile from python are assisting well on this behalf
# cv2.getTickCount() is like tic() toc() in other languages
e1 = cv2.getTickCount() # Is the same as time.time()
img = cv2.imread('test_image.png')
plt.imshow(img[:,:,::1])
plt.show()
e2 = cv2.getTickCount()
# cv2.getTickFrequency = numbers of clock-cycles per second
t = (e2 - e1) / cv2.getTickFrequency()
print("e2 - e1 = {}\n".format(e2 - e1))
print("t = {}".format(t))
# OpenCV has much optimized code by default:
cv2.setUseOptimized(True)
cv2.useOptimized()
%timeit res = cv2.medianBlur(img,49)
# With disabled optimization
cv2.setUseOptimized(False)
cv2.useOptimized()
%timeit res = cv2.medianBlur(img,49)
import cv2 as cv
import numpy as np
# 왜 최적화가 필요한가? 이미지처리에는 엄청나게 많은 연산이 진행되기 때문
# 배울 것
# cv.getTickCount
# cv.getTickFrequency
img = cv.imread("../1_arithmetic_operation/apple.jpg")
e1 = cv.getTickCount() # 코드 실행 시간
for i in range(5, 49, 2):
img = cv.medianBlur(img, i)
e2 = cv.getTickCount() # 코드 실행 종료 시간
t = (e2 - e1) * 1000 / cv.getTickFrequency() # 총 코드 실행 시간
print(t, "ms")
def getCMarkers (img):
e1 = cv2.getTickCount()
markers = []
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img3 = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 15, 10)
kernel = np.ones((2,2),np.uint8)
img2 = cv2.morphologyEx(img3, cv2.MORPH_CLOSE, kernel)
kernel = np.ones((2,2),np.uint8)
img2 = cv2.dilate(img2,kernel, 1)
# cv2.imshow("test1", img2)
# Setup SimpleBlobDetector parameters.
params = cv2.SimpleBlobDetector_Params()
params.filterByInertia = False
params.filterByConvexity = False
# Filter by Area.
params.filterByArea = True
params.minArea = 5
params.minDistBetweenBlobs = 1
# Filter by Circularity
params.filterByCircularity = True
params.minCircularity = 0.5
# # Filter by Convexity
params.filterByConvexity = True
params.minConvexity = 0.70
# Filter by Inertia
params.filterByInertia = True
params.minInertiaRatio = 0.2
params.filterByColor = False
# Create a detector with the parameters
detector = cv2.SimpleBlobDetector_create(params)
# Detect blobs.
keypoints = detector.detect(img2)
points = []
validpoints = []
for point in keypoints:
count = []
for p in keypoints:
distance = (p.pt[0]-point.pt[0])**2+(p.pt[1]-point.pt[1])**2
if p == point:
continue
elif distance <= (point.size*4.5)**2 and (abs(point.size/p.size-1) <= 0.3):
count.append(p.pt)
if len(count) >= 2:
points.append((point.pt, count))
validpoints.append(point)
for point in points:
p, near = point
# distance open the angre and 90 degree
midistance = math.pi/30.0
bottom = []
rigth = []
for p1 in near:
for p2 in near:
if p1 == p2:
continue
u = np.array([p1[0]-p[0], p1[1]-p[1]])
v = np.array([p2[0]-p[0], p2[1]-p[1]])
angle = np.math.atan2(np.linalg.det([u,v]),np.dot(u,v))
if abs(angle-math.pi/2.0) < math.pi/30.0:
bottom = p1
rigth = p2
conner = rigth+u
addu = u/6.0
addv = v/6.0
conners = [p-addu-addv, bottom+addu-addv, rigth-addu+addv, conner+addu+addv]
trans = transformMatrixPoints(conners, [10, 10], 10)
code = cv2.warpPerspective(gray, trans, dsize=(100, 100))
# code = cv2.adaptiveThreshold(code, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 25, 1)
number = getNumber(code, 160)
if number == False:
continue
# cv2.imshow("m2", code)
uu = np.array([0, 1])
angle = np.math.atan2(np.linalg.det([v,uu]),np.dot(v,uu))
mid = p+u*0.5+v*0.5
if number != 0:
markers.append([number, mid, angle])
# img2 = cv2.drawKeypoints(img2, validpoints, np.array([]), (0,0,255), cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
# cv2.imshow("m3", img2)
return markers
bbox = (287, 23, 86, 320)
# Uncomment the line below to select a different bounding box
bbox = cv2.selectROI(frame, False)
# Initialize tracker with first frame and bounding box
ok = tracker.init(frame, bbox)
while True:
# Read a new frame
ok, frame = video.read()
if not ok:
break
# Start timer
timer = cv2.getTickCount()
# Update tracker
ok, bbox = tracker.update(frame)
#fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer);
# Draw bounding box
if ok:
# Tracking success
p1 = (int(bbox[0]), int(bbox[1]))
p2 = (int(bbox[0] + bbox[2]), int(bbox[1] + bbox[3]))
cv2.rectangle(frame, p1, p2, (255, 0, 0), 2, 1)
else:
# Tracking failure
cv2.putText(frame, "Tracking failure detected", (100, 80),
def cascade_vh(self):
face_flag = 0
target_lose_cnt = 0
while (cv2.waitKey(1) != 27):
start_tc = cv2.getTickCount()
grabbed, frame = self.cap.read()
if not grabbed:
logging.info("READ OVER")
break
if not face_flag:
face_bbox = self.detector.face_detection(frame)
# No face has been detected
if isinstance(face_bbox, int):
cv2.imshow("frame", frame)
continue
# detect successfully
else:
logging.info("tracker init")
face_flag = 1
face_bbox = self.expand_bbox(*face_bbox)
self.tracker.start_track(frame, *face_bbox)
target_lose_cnt = 0
else:
# track
if target_lose_cnt < self.lose_threshold:
(success, box_predict) = self.tracker.update_track(frame)
if not success:
logging.debug("update failed")
target_lose_cnt += 1
cv2.imshow("frame", frame)
continue
(old_x, old_y, old_w, old_h) = (int(v)
for v in box_predict)
# draw predict rect
# self.draw_rect(frame,(old_x, old_y, old_w, old_h),(255,0,0))
# face_bbox is relative coordinates!!!!
face_bbox = self.detector.face_detection(
frame[old_y:old_y + old_h, old_x:old_x + old_w])
if isinstance(face_bbox, int):
target_lose_cnt += 1
cv2.imshow("frame", frame)
continue
target_lose_cnt = 0
face_bbox = self.expand_bbox(*face_bbox)
face_bbox[0] += int(old_x)
face_bbox[1] += int(old_y)
# lose target
else:
logging.info("losing target")
# reset tracker
self.tracker = Tracker()
face_flag = 0
cv2.imshow("frame", frame)
continue
# self.draw_rect(frame,face_bbox,(0,255,0))
face_rect = self.bbox_to_rect(face_bbox)
success, frame = self.run_face_swap(frame, face_rect)
if not success:
pass
cv2.imshow("output", frame)
end_tc = cv2.getTickCount()
fps = cv2.getTickFrequency() / (end_tc - start_tc)
logging.info("fps {}".format(fps))
cv2.destroyAllWindows()
self.cap.release()
# # frame = cv.pyrMeanShiftFiltering(frame, 20, 100)
# frame = cv.GaussianBlur(frame, (3, 3), 0)
# # hsv = cv.cvtColor(frame, cv.COLOR_BGR2HSV)
# # mask1 = cv.inRange(hsv, lowerb=lower_hsv, upperb=upper_hsv)
# mask = cv.inRange(frame, lowerb=np.array([0, 0, 110]), upperb=np.array([110, 110, 255]))
# kernel = cv.getStructuringElement(cv.MORPH_RECT, (3, 3))
# mask = cv.morphologyEx(mask, cv.MORPH_OPEN, kernel)
# # mask = cv.bitwise_or(mask1, mask2)
# # dst = cv.bitwise_and(frame, frame, mask=mask)
# contours, heriachy = cv.findContours(mask, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE)
# for i, contour in enumerate(contours):
# contour_area = cv.contourArea(contour)
# x, y, w, h = cv.boundingRect(contour)
# cv.drawContours(mask, contours, i, (0, 0, 255), 3)
# if contour_area > 0.8 * w * h:
# cv.drawContours(frame, contours, i, (255, 0, 255), 2)
# cv.imshow("mask", mask)
# cv.imshow("video", frame)
# if cv.waitKey(40) == 27:
# break
t1 = cv.getTickCount()
extrace_object_demo()
t2 = cv.getTickCount()
print("time: %s ms" % ((t2 - t1) / cv.getTickFrequency() * 1000))
# cv.destroyAllWindows()
from __future__ import print_function
import cv2 as cv
capture = cv.VideoCapture(0)
ret, frame_first = capture.read()
ret, frame_first = capture.read()
ret, frame_first = capture.read()
ret, frame_first = capture.read()
ret, frame_first = capture.read()
frame_first_gray = cv.cvtColor(frame_first, cv.COLOR_BGR2GRAY)
frame_first_blurred = cv.GaussianBlur(frame_first_gray, (5, 5), 0)
while True:
start = cv.getTickCount()
_, frame = capture.read()
if frame is None:
break
frame_gray = cv.cvtColor(frame, cv.COLOR_BGR2GRAY)
frame_blurred = cv.GaussianBlur(frame_gray, (5, 5), 0)
# frame_canny = cv.Canny(frame_blurred, 50, 150)
difference = cv.absdiff(frame_first_blurred, frame_blurred)
_, difference = cv.threshold(difference, 25, 255, cv.THRESH_BINARY)
difference_blurred = cv.GaussianBlur(difference, (5, 5), 0)
cv.imshow('BG Filtered', difference_blurred)
keyboard = cv.waitKey(30)
if keyboard == 'q' or keyboard == 27:
while True:
ok, frame = video.read()
if not ok: break
cv.imshow('Tracking', frame)
if cv.waitKey(1) != -1: break
bbox = cv.selectROI('Tracking', frame, showCrosshair=False)
print(bbox)
# Initialize tracker with first frame and bounding box
ok = tracker.init(frame, bbox)
print('tracker.init :', ok)
count = 0
fps = 0
t0 = cv.getTickCount()
while True:
ok, frame = video.read()
if not ok: break
if cv.waitKey(1) == 27: break
# Update tracker
ok, bbox = tracker.update(frame)
# Draw bounding box
if ok:
# Tracking success
p1 = (int(bbox[0]), int(bbox[1]))
p2 = (int(bbox[0] + bbox[2]), int(bbox[1] + bbox[3]))
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Human face detect and Track')
parser.add_argument('--video_path', default=0, help='Path for video')
args = parser.parse_args()
video_path = args.parse_args()
d = Detector()
t = Tracker()
cap = cv2.VideoCapture(video_path)
face_flag = 0
lose_threshold = 10
target_lose_cnt = 0
while (cv2.waitKey(1) != 27):
start_tc = cv2.getTickCount()
grabbed, frame = cap.read()
if not grabbed:
break
if not face_flag:
face_bbox = d.face_detection(frame)
# No face has been detected
if isinstance(face_bbox, int):
cv2.imshow("output", frame)
continue
# detect successfully
else:
logging.info("tracker init")
face_flag = 1
face_bbox = expand_bbox(*face_bbox)
SIZE = h_img.size
NUM_BIN = 256
n_threads = int(numpy.ceil((SIZE + NUM_BIN - 1) / NUM_BIN))
start = drv.Event()
end = drv.Event()
start.record()
# Calling kernel
atomic_hist(drv.Out(h_result),
drv.In(h_a),
numpy.uint32(SIZE),
block=(n_threads, 1, 1),
grid=(NUM_BIN, 1))
end.record()
end.synchronize()
secs = start.time_till(end) * 1e-3
print("Time for Calculating Histogram on GPU without shared memory")
print("%fs" % (secs))
plt.stem(h_result)
plt.xlim([0, 256])
plt.title("Histogram on GPU")
start = cv2.getTickCount()
# OpenCV calcHist
hist = cv2.calcHist([h_img], [0], None, [256], [0, 256])
end = cv2.getTickCount()
secs = (end - start) / cv2.getTickFrequency()
print("Time for Calculating Histogram using OpenCV")
print("%fs" % (secs))
# plt.stem(hist)
# plt.xlim([0,256])
def main(model_code):
# create tracker
tracker_info = Tracker(args.tracker_name, args.tracker_param, None)
params = tracker_info.get_parameters()
params.visualization = args.vis
params.debug = args.debug
params.visdom_info = {
'use_visdom': False,
'server': '127.0.0.1',
'port': 8097
}
tracker = tracker_info.tracker_class(params)
RF_module = RefineModule(refine_path.format(model_code),
selector_path,
search_factor=sr,
input_sz=input_sz)
model_name = args.tracker_name + '_' + args.tracker_param +\
'{}-{}'.format(RF_type.format(model_code), selector_path) + '_%d'%(args.run_id)
# create dataset
dataset = DatasetFactory.create_dataset(name=args.dataset,
dataset_root=dataset_root_,
load_img=False)
if args.dataset in ['VOT2016', 'VOT2018', 'VOT2019']:
total_lost = 0
# restart tracking
for v_idx, video in enumerate(dataset):
if args.video != '':
# test one special video
if video.name != args.video:
continue
frame_counter = 0
lost_number = 0
toc = 0
'''对refinement module计时'''
toc_refine = 0
pred_bboxes = []
for idx, (img, gt_bbox) in enumerate(video):
if len(gt_bbox) == 4:
gt_bbox = [
gt_bbox[0], gt_bbox[1], gt_bbox[0],
gt_bbox[1] + gt_bbox[3] - 1,
gt_bbox[0] + gt_bbox[2] - 1,
gt_bbox[1] + gt_bbox[3] - 1,
gt_bbox[0] + gt_bbox[2] - 1, gt_bbox[1]
]
tic = cv2.getTickCount()
'''get RGB format image'''
img_RGB = img[:, :, ::-1].copy() # BGR --> RGB
if idx == frame_counter:
H, W, _ = img.shape
cx, cy, w, h = get_axis_aligned_bbox(np.array(gt_bbox))
gt_bbox_ = [cx - (w - 1) / 2, cy - (h - 1) / 2, w, h]
'''Initialize'''
gt_bbox_np = np.array(gt_bbox_)
gt_bbox_torch = torch.from_numpy(
gt_bbox_np.astype(np.float32))
init_info = {}
init_info['init_bbox'] = gt_bbox_torch
_ = tracker.initialize(img_RGB, init_info)
'''##### initilize refinement module for specific video'''
RF_module.initialize(img_RGB, np.array(gt_bbox_))
pred_bbox = gt_bbox_
pred_bboxes.append(1)
elif idx > frame_counter:
'''Track'''
outputs = tracker.track(img_RGB)
pred_bbox = outputs['target_bbox']
'''##### refine tracking results #####'''
pred_bbox = RF_module.refine(img_RGB, np.array(pred_bbox))
'''report result and update state'''
x1, y1, w, h = pred_bbox.tolist()
'''add boundary and min size limit'''
x1, y1, x2, y2 = bbox_clip(x1, y1, x1 + w, y1 + h, (H, W))
w = x2 - x1
h = y2 - y1
new_pos = torch.from_numpy(
np.array([y1 + h / 2, x1 + w / 2]).astype(np.float32))
new_target_sz = torch.from_numpy(
np.array([h, w]).astype(np.float32))
new_scale = torch.sqrt(new_target_sz.prod() /
tracker.base_target_sz.prod())
##### update
tracker.pos = new_pos.clone()
tracker.target_sz = new_target_sz
tracker.target_scale = new_scale
overlap = vot_overlap(pred_bbox, gt_bbox,
(img.shape[1], img.shape[0]))
if overlap > 0:
# not lost
pred_bboxes.append(pred_bbox)
else:
# lost object
pred_bboxes.append(2)
frame_counter = idx + 5 # skip 5 frames
lost_number += 1
else:
pred_bboxes.append(0)
toc += cv2.getTickCount() - tic
if idx == 0:
cv2.destroyAllWindows()
if args.vis and idx > frame_counter:
cv2.polylines(
img, [np.array(gt_bbox, np.int).reshape(
(-1, 1, 2))], True, (0, 255, 0), 3)
if len(pred_bbox) == 8:
cv2.polylines(
img,
[np.array(pred_bbox, np.int).reshape(
(-1, 1, 2))], True, (0, 255, 255), 3)
else:
bbox = list(map(int, pred_bbox))
cv2.rectangle(img, (bbox[0], bbox[1]),
(bbox[0] + bbox[2], bbox[1] + bbox[3]),
(0, 255, 255), 3)
cv2.putText(img, str(idx), (40, 40),
cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 255), 2)
cv2.putText(img, str(lost_number), (40, 80),
cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)
cv2.imshow(video.name, img)
cv2.waitKey(1)
toc /= cv2.getTickFrequency()
# save results
video_path = os.path.join(save_dir, args.dataset, model_name,
'baseline', video.name)
if not os.path.isdir(video_path):
os.makedirs(video_path)
result_path = os.path.join(video_path,
'{}_001.txt'.format(video.name))
with open(result_path, 'w') as f:
for x in pred_bboxes:
if isinstance(x, int):
f.write("{:d}\n".format(x))
else:
f.write(','.join([vot_float2str("%.4f", i)
for i in x]) + '\n')
print(
'({:3d}) Video: {:12s} Time: {:4.1f}s Speed: {:3.1f}fps Lost: {:d}'
.format(v_idx + 1, video.name, toc, idx / toc, lost_number))
total_lost += lost_number
print("{:s} total lost: {:d}".format(model_name, total_lost))
import cv2
face_cascade = cv2.CascadeClassifier("./haarcascade_frontalface_alt2.xml")
cap = cv2.VideoCapture(0)
while True:
ret, frame = cap.read()
tickmark = cv2.getTickCount()
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
face = face_cascade.detectMultiScale(gray, scaleFactor=1.2, minNeighbors=3)
for x, y, w, h in face:
cv2.rectangle(frame, (x, y), (x + w, y + h), (255, 0, 0), 2)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
fps = cv2.getTickFrequency() / (cv2.getTickCount() - tickmark)
cv2.putText(frame, "FPS: {:05.2f}".format(fps), (10, 30),
cv2.FONT_HERSHEY_PLAIN, 2, (255, 0, 0), 2)
cv2.imshow('video', frame)
cap.release()
cv2.destroyAllWindows()
import cv2
cv2.setUseOptimized(True)
e1 = cv2.getTickCount()
img1 = cv2.imread('capture 0.png')
img2 = cv2.imread('images.png')
rows, cols, channels = img2.shape
roi = img1[0:rows, 0:cols]
img2gray = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
ret, mask = cv2.threshold(img2gray, 1, 10, cv2.THRESH_BINARY)
mask_inv = cv2.bitwise_not(mask)
img1_bg = cv2.bitwise_and(roi, roi, mask=mask_inv)
img2_fg = cv2.bitwise_and(img2, img2, mask=mask)
dst = cv2.add(img1_bg, img2_fg)
img1[0:rows, 0:cols] = dst
cv2.imshow('image', img1)
e2 = cv2.getTickCount()
t = (e2 - e1) / cv2.getTickFrequency()
print t
cv2.waitKey(0)
def debug(msg):
if Debug.debug_mode:
print("%f: %s" % (
cv.getTickCount(),
msg
))
# programa para calcular frame fps en varios escenarios
# 1) basico sin sin ninguna programacion especial.
# prom 48 fps, min 34, max 68.
import cv2
captura = cv2.VideoCapture(0)
while (captura.isOpened()):
timer = cv2.getTickCount() # inicia el tiempo de fps
ret, imagen = captura.read()
if ret == True:
fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer
) # calcula el tiempo fps
# voltea la imagen
imagen = cv2.flip(imagen, 0)
cv2.putText(imagen, str(int(fps)), (75, 50), cv2.FONT_HERSHEY_COMPLEX,
0.7, (0, 255, 0), 2)
cv2.imshow('video', imagen)
#print("Frames por segundos: ", str(int(fps)))
if cv2.waitKey(1) & 0xFF == ord('s'):
break
else:
break
captura.release()
cv2.destroyAllWindows()
return data, rows, cols, dims
train_data, rows, cols, dims = image_read(
'D:\\Learning materials\\machine_homework\\training4'
) # 载入对应的训练图片,共20个训练样本
test_data = cv.imread(
'D:\\Learning materials\\machine_homework\\test3\\image21.jpg'
) # 载入对应的测试图片，有一个测试样本
test_data = np.asarray(test_data) # 将图片转化成数组形式
h, temp = 100, 1 # 带宽值设定为120，中间变量为1
P = np.zeros((rows, cols)) # 一个二维的全0数组,为最后的输出概率
P1 = np.empty((rows, cols, dims)) # 三维的全空数组
P2 = np.zeros((rows, cols)) # 二维的全0数组
t1 = cv.getTickCount() # 计算整个kde算法所需要的时间
for i in range(19): # 共19个训练样本
for j in range(rows):
for k in range(cols):
for l in range(dims):
P1[j, k, l] = 1 - pow(
(test_data[j, k, l] - train_data[j, k, l, i]) / h, 2)
if P1[j, k, l] < 0: # 如果概率密度函数值小于0，那么将其赋值为0
P1[j, k, l] = 0
temp = temp * P1[j, k, l]
P2[j, k] = temp
temp = 1.0
P = P + P2 # 概率值叠加
P = 15 / (8 * pi * 20 * h**3) * P # EP函数的表达式，其中N=20