def app(image_path):
labels = load_labels(cfg.LABEL_PATH)
interpreter = Interpreter(cfg.DETECT_MODEL_PATH)
interpreter.allocate_tensors()
_, input_H, input_W, _ = interpreter.get_input_details()[0]['shape']
for i in range(5):
image = cv.imread(image_path)
assert image is not None, 'Invalid image path'
_start_t = time.time()
inp_image = cv.cvtColor(image, cv.COLOR_BGR2RGB)
inp_image = cv.resize(inp_image, (input_W, input_H))
scores, class_ids, bboxes = detect_objects(interpreter,
inp_image, image.shape[:2], thresh=0.6)
_prx_t = time.time() - _start_t
LOG.info('FPS: {:.3f}'.format(1/_prx_t))
image = vis.plotBBoxes(image, bboxes.astype('int'),
classes=[labels[idx] for idx in class_ids],
scores=scores)
image = vis.plotInfo(image, 'Raspberry Pi - FPS: {:.2f}'.format(1/_prx_t))
image = cv.cvtColor(np.asarray(image), cv.COLOR_BGR2RGB)
cv.imwrite('output.jpg', image)
def __drawBoxes(self):
if len(self.detectedObjects):
labels = len(self.detectedObjects) * ['strawberry']
confidence = len(self.detectedObjects) * [0]
coordinates = [(x, y, x + w, y + h) for x, y, w, h in self.detectedObjects]
self.frame = vis.plotBBoxes(self.frame, coordinates, labels, confidence)
self.frame = vis.plotInfo(self.frame, 'Flint View')
self.frame = cv.cvtColor(np.asarray(self.frame), cv.COLOR_BGR2RGB)
def app(video_link, video_name, show, record, flip_hor, flip_ver):
# initialize Face Detection net
face_detector = FaceDetector()
# initialize Emotioner net
emotioner = Emotioner()
# initialize Video Capturer
cap = cv.VideoCapture(video_link)
(W, H), FPS = imgproc.cameraCalibrate(cap)
LOG.info('Camera Info: ({}, {}) - {:.3f}'.format(W, H, FPS))
if record:
time_str = time.strftime(cfg.TIME_FM)
writer = cv.VideoWriter(video_name+time_str+'.avi',
cv.VideoWriter_fourcc(*'XVID'), FPS, (W, H))
while cap.isOpened():
_, frm = cap.read()
if not _:
LOG.info('Reached the end of Video source')
break
if flip_ver: frm = cv.flip(frm, 0)
if flip_hor: frm = cv.flip(frm, 1)
_start_t = time.time()
scores, bboxes = face_detector.getFaces(frm, def_score=0.5)
emos = []
pbboxes = []
for i, bbox in enumerate(bboxes):
x1, y1, x2, y2 = getPadding(frm, bbox)
face_img = frm[y1:y2, x1:x2]
emo_idx = emotioner.getEmotion(face_img)
emos.append(EMOTION[emo_idx])
pbboxes.append((x1, y1, x2, y2))
_prx_t = time.time() - _start_t
if len(bboxes):
frm = vis.plotBBoxes(frm, pbboxes, emos)
frm = vis.plotInfo(frm, 'Raspberry Pi - FPS: {:.3f}'.format(1/_prx_t))
frm = cv.cvtColor(np.asarray(frm), cv.COLOR_BGR2RGB)
if record:
writer.write(frm)
if show:
cv.imshow(video_name, frm)
key = cv.waitKey(1)
if key in [27, ord('q')]:
LOG.info('Interrupted by Users')
break
if record:
writer.release()
cap.release()
cv.destroyAllWindows()
def app(video_link, video_name, show, flip_hor, flip_ver):
# initialize Face Detection net
face_detector = FaceDetector()
LOG.info('Face Detector initialization done')
# initialize Face Landmark net
face_landmarker = FaceLandmarker()
LOG.info('Face Landmarker initialization done')
# initialize Video Capturer
cap = cv.VideoCapture(video_link)
(W, H), FPS = imgproc.cameraCalibrate(cap, size=720, by_height=True)
LOG.info('Camera Info: ({}, {}) - {:.3f}'.format(W, H, FPS))
while cap.isOpened():
_, frm = cap.read()
if not _:
LOG.info('Reached the end of Video source')
break
if flip_ver: frm = cv.flip(frm, 0)
if flip_hor: frm = cv.flip(frm, 1)
frm = imgproc.resizeByHeight(frm, 720)
_start_t = time.time()
scores, bboxes = face_detector.getFaces(frm, def_score=0.5)
landmarks = []
for i, bbox in enumerate(bboxes):
x1, y1, x2, y2 = bbox
face_img = frm[y1:y2, x1:x2]
landmark = face_landmarker.getLandmark(face_img)
landmark[:, :] += np.array([x1, y1])
landmarks.append(landmark)
_prx_t = time.time() - _start_t
if len(bboxes):
for i, landmark in enumerate(landmarks):
for j, point in enumerate(landmark):
cv.circle(frm, tuple(point), 3, (0, 255, 0), -1)
frm = vis.plotBBoxes(frm, bboxes, len(bboxes) * ['face'], scores)
frm = vis.plotInfo(frm,
'Raspberry Pi - FPS: {:.3f}'.format(1 / _prx_t))
frm = cv.cvtColor(np.asarray(frm), cv.COLOR_BGR2RGB)
if show:
cv.imshow(video_name, frm)
key = cv.waitKey(1)
if key in [27, ord('q')]:
LOG.info('Interrupted by Users')
break
cap.release()
cv.destroyAllWindows()
def app(video_link, video_name, show=False, record=False):
# initialize TFLite model
labels = load_labels(cfg.LABEL_PATH)
interpreter = Interpreter(cfg.DETECT_MODEL_PATH)
interpreter.allocate_tensors()
_, input_H, input_W, _ = interpreter.get_input_details()[0]['shape']
# initialize video capturer
cap = cv.VideoCapture(video_link)
(W, H), FPS = imgproc.cameraCalibrate(cap)
FRM_MOD = int(FPS / cfg.pFPS + 0.5)
LOG.info('Camera Info: ({}, {}) - {:.3f}'.format(W, H, FPS))
if record:
writer = imgproc.getVideoRecorder(video_name, (W, H), FPS)
frm_cnt = 0
while cap.isOpened():
_, frm = cap.read()
if not _:
LOG.info('Reached the end of Video source')
break
frm_cnt += 1
if frm_cnt % FRM_MOD: continue
_start_t = time.time()
inp_image = cv.cvtColor(frm, cv.COLOR_BGR2RGB)
inp_image = cv.resize(inp_image, (input_W, input_H))
scores, class_ids, bboxes = detect_objects(interpreter,
inp_image, frm.shape[:2], thresh=0.6)
_prx_t = time.time() - _start_t
LOG.info('FPS: {:.3f}'.format(1/_prx_t))
frm = vis.plotBBoxes(frm, bboxes.astype('int'),
classes=[labels[idx] for idx in class_ids],
scores=scores)
frm = vis.plotInfo(frm, 'Raspberry Pi - FPS: {:.2f}'.format(1/_prx_t))
frm = cv.cvtColor(np.asarray(frm), cv.COLOR_BGR2RGB)
if record:
writer.write(frm)
if show:
cv.imshow(video_name, frm)
key = cv.waitKey(1)
if key in [27, ord('q')]:
LOG.info('Interrupted by Users')
break
if record:
writer.release()
cap.release()
cv.destroyAllWindows()
def app(image_path):
# initialize Face Detection net
face_detector = FaceDetector()
LOG.info('Face Detector initialization done')
# initialize Face Landmark net
face_landmarker = FaceLandmarker()
LOG.info('Face Landmarker initialization done')
cap = cv.VideoCapture(0)
while cap.isOpened():
_, image = cap.read()
if not _:
LOG.info('Reached the end of Video source')
break
image = imgproc.resizeByHeight(image, 720)
_start_t = time.time()
scores, bboxes = face_detector.getFaces(image, def_score=0.5)
landmarks = []
for i, bbox in enumerate(bboxes):
x1, y1, x2, y2 = bbox
face_img = image[y1:y2, x1:x2]
landmark = face_landmarker.getLandmark(face_img)
aligned_img = alignFace(face_img, landmark)
cv.imshow('aligned-faces' + str(i), aligned_img)
landmark[:, :] += np.array([x1, y1])
landmarks.append(landmark)
_prx_t = time.time() - _start_t
if len(bboxes):
for i, landmark in enumerate(landmarks):
for j, point in enumerate(landmark):
cv.circle(image, tuple(point), 3, (0, 255, 0), -1)
image = vis.plotBBoxes(image, bboxes,
len(bboxes) * ['face'], scores)
image = vis.plotInfo(image,
'Raspberry Pi - FPS: {:.3f}'.format(1 / _prx_t))
image = cv.cvtColor(np.asarray(image), cv.COLOR_BGR2RGB)
cv.imshow(image_path, image)
key = cv.waitKey(1)
if key in [27, ord('q')]:
LOG.info('Interrupted by Users')
break
cap.release()
cv.destroyAllWindows()
def app(video_link, video_name, show, record, flip_hor, flip_ver):
# initialize Face Detection net
object_detector = ObjectDetector()
# initialize Video Capturer
cap = cv.VideoCapture(video_link)
(W, H), FPS = imgproc.cameraCalibrate(cap)
LOG.info('Camera Info: ({}, {}) - {:.3f}'.format(W, H, FPS))
if record:
time_str = time.strftime(cfg.TIME_FM)
writer = cv.VideoWriter(video_name + time_str + '.avi',
cv.VideoWriter_fourcc(*'XVID'), 20,
(1280, 720))
cnt_frm = 0
while cap.isOpened():
_, frm = cap.read()
if not _:
LOG.info('Reached the end of Video source')
break
cnt_frm += 1
if flip_ver: frm = cv.flip(frm, 0)
if flip_hor: frm = cv.flip(frm, 1)
frm = imgproc.resizeByHeight(frm, 720)
_start_t = time.time()
scores, bboxes = object_detector.getObjects(frm, def_score=0.5)
_prx_t = time.time() - _start_t
if len(bboxes):
frm = vis.plotBBoxes(frm, bboxes, len(bboxes) * ['person'], scores)
frm = vis.plotInfo(frm,
'Raspberry Pi - FPS: {:.3f}'.format(1 / _prx_t))
frm = cv.cvtColor(np.asarray(frm), cv.COLOR_BGR2RGB)
if record:
writer.write(frm)
if show:
cv.imshow(video_name, frm)
key = cv.waitKey(1)
if key in [27, ord('q')]:
LOG.info('Interrupted by Users')
break
if record:
writer.release()
cap.release()
cv.destroyAllWindows()
def app(image_path):
# initialize Face Detection net
face_detector = FaceDetector()
LOG.info('Face Detector initialization done')
# initialize Face Landmark net
face_landmarker = FaceLandmarker()
LOG.info('Face Landmarker initialization done')
# initialize Face Alignment class
face_aligner = FaceAligner()
LOG.info('Face Aligner initialization done')
# initialize Video Capturer
image = cv.imread(image_path)
assert image is not None
image = imgproc.resizeByHeight(image, 720)
_start_t = time.time()
scores, bboxes = face_detector.getFaces(image, def_score=0.5)
landmarks = []
for i, bbox in enumerate(bboxes):
x1, y1, x2, y2 = bbox
face_img = image[y1:y2, x1:x2]
landmark = face_landmarker.getLandmark(face_img)
aligned_img = face_aligner.align(face_img, landmark)
cv.imshow('aligned-faces' + str(i), aligned_img)
landmark[:, :] += np.array([x1, y1])
landmarks.append(landmark)
_prx_t = time.time() - _start_t
if len(bboxes):
for i, landmark in enumerate(landmarks):
for j, point in enumerate(landmark):
cv.circle(image, tuple(point), 3, (0, 255, 0), -1)
image = vis.plotBBoxes(image, bboxes, len(bboxes) * ['face'], scores)
image = vis.plotInfo(image,
'Raspberry Pi - FPS: {:.3f}'.format(1 / _prx_t))
image = cv.cvtColor(np.asarray(image), cv.COLOR_BGR2RGB)
cv.imshow(image_path, image)
key = cv.waitKey(0)
cv.destroyAllWindows()
def app(video_link, video_name, show, flip_hor, flip_ver):
# initialize Face Detection net
face_detector = FaceDetector()
# initialize Video Capturer
cap = cv.VideoCapture(video_link)
(W, H), FPS = imgproc.cameraCalibrate(cap)
LOG.info('Camera Info: ({}, {}) - {:.3f}'.format(W, H, FPS))
while cap.isOpened():
_, frm = cap.read()
if not _:
LOG.info('Reached the end of Video source')
break
if flip_ver: frm = cv.flip(frm, 0)
if flip_hor: frm = cv.flip(frm, 1)
_start_t = time.time()
scores, bboxes = face_detector.getFaces(frm, def_score=0.5)
_prx_t = time.time() - _start_t
if len(bboxes):
frm = vis.plotBBoxes(frm, bboxes, len(bboxes) * ['face'], scores)
frm = vis.plotInfo(frm,
'Raspberry Pi - FPS: {:.3f}'.format(1 / _prx_t))
frm = cv.cvtColor(np.asarray(frm), cv.COLOR_BGR2RGB)
if show:
cv.imshow(video_name, frm)
key = cv.waitKey(1)
if key in [27, ord('q')]:
LOG.info('Interrupted by Users')
break
cap.release()
cv.destroyAllWindows()
def processFrame(frm, face_detector, face_landmarker):
_start_t = time.time()
scores, bboxes = face_detector.getFaces(frm, def_score=0.5)
landmarks = []
for i, bbox in enumerate(bboxes):
x1, y1, x2, y2 = bbox
face_img = frm[y1:y2, x1:x2]
landmark = face_landmarker.getLandmark(face_img)
landmark[:, :] += np.array([x1, y1])
landmarks.append(landmark)
_prx_t = time.time() - _start_t
if len(bboxes):
for i, landmark in enumerate(landmarks):
for j, point in enumerate(landmark):
cv.circle(frm, tuple(point), 3, (0, 255, 0), -1)
frm = vis.plotBBoxes(frm, bboxes, len(bboxes) * ['face'], scores)
frm = vis.plotInfo(frm, 'Raspberry Pi - FPS: {:.3f}'.format(1 / _prx_t))
frm = cv.cvtColor(np.asarray(frm), cv.COLOR_BGR2RGB)
return frm
continue
start_time = time.time()
elif (len(bboxes) == 0) and (time.time() - start_time) > delay:
tracking = 0
stage = SEARCH_LR
#go home call
my_robot.goReady()
print("go home,cant see anything")
start_time = time.time()
if len(bboxes):
frm = vis.plotBBoxes(frm,
[(x, y, x + w, y + h) for x, y, w, h in bboxes],
len(bboxes) * ['strawberry'],
len(bboxes) * [0])
frm = vis.plotInfo(frm, 'Raspberry Pi - FPS: {:.3f}'.format(1 / _prx_t))
frm = cv.cvtColor(np.asarray(frm), cv.COLOR_BGR2RGB)
# show the frame
cv.imshow("Frame", frm)
cv.imwrite("../Frontend/frame.jpg", frm)
key = cv.waitKey(1)
if key in [27, ord('q')]:
break
#time.sleep(2)
pipeline.stop()
def app(video_link, video_name, show, record, flip_hor, flip_ver):
# initialize Face Detection net
object_detector = ObjectDetector()
# initialize Video Capturer
#cap = cv.VideoCapture(video_link)
cap = WebcamVideoStream(src=cfg.RTSP_URL).start()
# (W, H), FPS = imgproc.cameraCalibrate(cap.stream)
# LOG.info('Camera Info: ({}, {}) - {:.3f}'.format(W, H, FPS))
if record:
time_str = time.strftime(cfg.TIME_FM)
writer = cv.VideoWriter(video_name + time_str + '.avi',
cv.VideoWriter_fourcc(*'XVID'), 20,
(1280, 720))
#cnt_frm = 0
while True:
frm = cap.read()
if frm is None:
continue
#cnt_frm += 1
# if(cnt_frm % 5 != 0):
# continue
if flip_ver:
frm = cv.flip(frm, 0)
if flip_hor:
frm = cv.flip(frm, 1)
frm = imgproc.resizeByHeight(frm, 720)
_start_t = time.time()
images, bboxes = object_detector.getObjects(frm, def_score=0.5)
if images:
files = []
for im in images:
byte_io = BytesIO()
im.save(byte_io, 'png')
byte_io.seek(0)
files.append(('files', byte_io))
files.append(
('bboxes', (None, json.dumps(bboxes), 'application/json')))
if (len(bboxes)):
try:
requests.post(url=URL, files=files)
except:
continue
_prx_t = time.time() - _start_t
if len(bboxes):
frm = vis.plotBBoxes(frm, bboxes, len(bboxes) * ['person'])
frm = vis.plotInfo(frm,
'Raspberry Pi - FPS: {:.3f}'.format(1 / _prx_t))
frm = cv.cvtColor(np.asarray(frm), cv.COLOR_BGR2RGB)
if record:
writer.write(frm)
# if show:
# cv.imshow(video_name, frm)
# key = cv.waitKey(1)
# if key in [27, ord('q')]:
# LOG.info('Interrupted by Users')
# break
if record:
writer.release()
def app(video_link, video_name, show, record, flip_hor, flip_ver):
# initialize Face Detection net
config = read_py_config('config.py')
object_detector = ObjectDetector()
reid = PersonEmbedder()
# initialize Video Capturer
cap = MulticamCapture(video_link)
#cap = WebcamVideoStream(src=video_link).start()
# (W, H), FPS = imgproc.cameraCalibrate(cap)
# LOG.info('Camera Info: ({}, {}) - {:.3f}'.format(W, H, FPS))
tracker = MultiCameraTracker(cap.get_num_sources(), reid, **config)
thread_body = FramesThreadBody(cap, max_queue_length=len(cap.captures) * 2)
frames_thread = Thread(target=thread_body)
frames_thread.start()
if record:
time_str = time.strftime(cfg.TIME_FM)
writer = cv.VideoWriter(video_name + time_str + '.avi',
cv.VideoWriter_fourcc(*'XVID'), 20,
(1280, 720))
cnt_frm = 0
counter = 0
while thread_body.process:
try:
frm = thread_body.frames_queue.get_nowait()
except queue.Empty:
frm = None
if frm is None:
continue
cnt_frm += 1
# if flip_ver: frm = cv.flip(frm, 0)
# if flip_hor: frm = cv.flip(frm, 1)
_start_t = time.time()
all_detections = []
for f in frm:
f = imgproc.resizeByHeight(f, 640)
_, bboxes = object_detector.getObjects(f, def_score=0.5)
all_detections.append(bboxes)
tracker.process(frm, all_detections, [[]])
tracked_objects = tracker.get_tracked_objects()
_prx_t = time.time() - _start_t
fps = round(1 / _prx_t, 1)
if len(bboxes):
frm = visualize_multicam_detections(frm, tracked_objects, fps)
frm = vis.plotInfo(frm,
'Raspberry Pi - FPS: {:.3f}'.format(1 / _prx_t))
frm = cv.cvtColor(np.asarray(frm), cv.COLOR_BGR2RGB)
if record:
writer.write(frm)
if show:
cv.imshow(video_name, frm)
key = cv.waitKey(1)
if key in [27, ord('q')]:
LOG.info('Interrupted by Users')
break
else:
if (counter % 10 == 0):
print(
f"IN : {SingleCameraTracker.COUNT_IN}, OUT: {SingleCameraTracker.COUNT_OUT}"
)
counter += 1
thread_body.process = False
frames_thread.join()
def app(video_path, video_name, show=False, record=False):
detector = FaceDetector()
plugin = detector.getPlugin()
poser = HeadPoser(plugin=plugin)
cap = cv.VideoCapture(video_path)
(W, H), FPS = imgproc.cameraCalibrate(cap)
_, frm = cap.read()
frm = imgproc.resizeByHeight(frm)
H, W = frm.shape[:2]
FRM_MOD = int(1. * FPS / cfg.pFPS + 0.5)
LOG.info('Camera Info: ({}, {}) - {:.3f}'.format(W, H, FPS))
if record:
time_str = time.strftime(cfg.TIME_FM)
writer = cv.VideoWriter(video_name + time_str + '.avi',
cv.VideoWriter_fourcc(*'XVID'), FPS, (W, H))
cnt_frm = 0
while cap.isOpened():
_, frm = cap.read()
if not _:
LOG.info('Reached the end of Video source')
break
cnt_frm += 1
if cnt_frm % FRM_MOD: continue
print(cnt_frm)
frm = imgproc.resizeByHeight(frm)
_start_t = time.time()
scores, bboxes = detector.getFaces(frm)
if len(bboxes):
for i, bbox in enumerate(bboxes):
x1, y1, x2, y2 = bbox
face_image = frm[y1:y2, x1:x2]
yaw, pitch, roll = poser.estimatePose(face_image)
cpoint = [(x1 + x2) // 2, (y1 + y2) // 2]
frm = hvis.draw(frm, cpoint, (yaw, pitch, roll))
_prx_t = time.time() - _start_t
frm = vis.plotInfo(frm,
'Raspberry Pi - FPS: {:.3f}'.format(1 / _prx_t))
frm = cv.cvtColor(np.asarray(frm), cv.COLOR_BGR2RGB)
if record:
writer.write(frm)
if show:
cv.imshow(video_name, frm)
key = cv.waitKey(1)
if key in [27, ord('q')]:
LOG.info('Interrupted by Users')
break
if record:
writer.release()
cap.release()
cv.destroyAllWindows()
def app(video_link, video_name, show, record, flip_hor, flip_ver):
# initialize Face Detection net
config = read_py_config('config.py')
object_detector = ObjectDetector()
reid = PersonEmbedder()
# initialize Video Capturer
#cap = MulticamCapture(video_link)
cap = WebcamVideoStream(src=video_link).start()
(W, H), FPS = imgproc.cameraCalibrate(cap)
LOG.info('Camera Info: ({}, {}) - {:.3f}'.format(W, H, FPS))
tracker = MultiCameraTracker(1, reid, **config)
if record:
time_str = time.strftime(cfg.TIME_FM)
writer = cv.VideoWriter(video_name + time_str + '.avi',
cv.VideoWriter_fourcc(*'XVID'), 20,
(1280, 720))
cnt_frm = 0
counter = 0
while True:
frm = cap.read()
if frm is None:
continue
cnt_frm += 1
# if flip_ver: frm = cv.flip(frm, 0)
# if flip_hor: frm = cv.flip(frm, 1)
_start_t = time.time()
all_detections = []
frm = imgproc.resizeByHeight(frm, 640)
_, bboxes = object_detector.getObjects(frm, def_score=0.5)
all_detections.append(bboxes)
tracker.process([frm], all_detections, [[]])
tracked_objects = tracker.get_tracked_objects()
_prx_t = time.time() - _start_t
fps = round(1 / _prx_t, 1)
if len(bboxes):
frm = visualize_multicam_detections([frm], tracked_objects, fps)
frm = vis.plotInfo(frm,
'Raspberry Pi - FPS: {:.3f}'.format(1 / _prx_t))
frm = cv.cvtColor(np.asarray(frm), cv.COLOR_BGR2RGB)
if record:
writer.write(frm)
if show:
cv.imshow(video_name, frm)
key = cv.waitKey(1)
if key in [27, ord('q')]:
LOG.info('Interrupted by Users')
break
else:
if (counter % 10 == 0):
print(
f"IN : {SingleCameraTracker.COUNT_IN}, OUT: {SingleCameraTracker.COUNT_OUT}"
)
counter += 1
if record:
writer.release()
cap.release()
cv.destroyAllWindows()
