def detect(self, frame, **kwargs):
# convert roi-plugin to detector roi
roi = Roi(*self.g_pool.roi.bounds)
debug_img = frame.bgr if self.g_pool.display_mode == "algorithm" else None
result = self.detector_2d.detect(
gray_img=frame.gray,
color_img=debug_img,
roi=roi,
)
norm_pos = normalize(result["location"], (frame.width, frame.height),
flip_y=True)
# Create basic pupil datum
datum = self.create_pupil_datum(
norm_pos=norm_pos,
diameter=result["diameter"],
confidence=result["confidence"],
timestamp=frame.timestamp,
)
# Fill out 2D model data
datum["ellipse"] = {}
datum["ellipse"]["axes"] = result["ellipse"]["axes"]
datum["ellipse"]["angle"] = result["ellipse"]["angle"]
datum["ellipse"]["center"] = result["ellipse"]["center"]
return datum
def detect(self, frame):
roi = Roi(*self.g_pool.u_r.get()[:4])
if (not 0 <= roi.x_min <= roi.x_max < frame.width
or not 0 <= roi.y_min <= roi.y_max < frame.height):
# TODO: Invalid ROIs can occur when switching camera resolutions, because we
# adjust the roi only after all plugin recent_events() have been called.
# Optimally we make a plugin out of the ROI and call its recent_events()
# immediately after the backend, before the detection.
logger.debug(
f"Invalid Roi {roi} for img {frame.width}x{frame.height}!")
return None
debug_img = frame.bgr if self.g_pool.display_mode == "algorithm" else None
result = self.detector_2d.detect(
gray_img=frame.gray,
color_img=debug_img,
roi=roi,
)
eye_id = self.g_pool.eye_id
location = result["location"]
result["norm_pos"] = normalize(location, (frame.width, frame.height),
flip_y=True)
result["timestamp"] = frame.timestamp
result["topic"] = f"pupil.{eye_id}"
result["id"] = eye_id
result["method"] = "2d c++"
return result
def process(self, img):
if img is None:
return
height, width = img.shape[0], img.shape[1]
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
timestamp = uvc.get_time_monotonic()
roi = None
if self.bbox is not None:
xmin, ymin, w, h = self.bbox
roi = Roi(xmin, ymin, xmin+w, ymin+h)
result = self.detector.detect(gray, timestamp, roi=roi)
#print(result)
if result["model_confidence"] > 0.25:
sphere = result["projected_sphere"]
self.__draw_ellipse(sphere, img, (255,120,120), 1)
if result["confidence"] > 0.5:
n = np.array(result['circle_3d']['normal'])
self.bbox = self.__get_bbox(result, img)
self.__draw_tracking_info(result, img)
# cv2.imshow("testando", img)
# cv2.waitKey(1)
self.pos = np.array([n[0], n[1], n[2], time.monotonic()])
self.countdown = 5
else:
self.countdown -= 1
if self.countdown <= 0:
self.pos = None
self.bbox = None
return img
def detect(self, frame):
roi = Roi(*self.g_pool.u_r.get()[:4])
result = self.detector_2d.detect(gray_img=frame.gray,
color_img=frame.bgr,
roi=roi)
eye_id = self.g_pool.eye_id
location = result["location"]
result["norm_pos"] = normalize(location, (frame.width, frame.height),
flip_y=True)
result["timestamp"] = frame.timestamp
result["topic"] = f"pupil.{eye_id}"
result["id"] = eye_id
result["method"] = "2d c++"
return result
def detect(self, frame, **kwargs):
# convert roi-plugin to detector roi
roi = Roi(*self.g_pool.roi.bounds)
debug_img = frame.bgr if self.g_pool.display_mode == "algorithm" else None
result = self.detector_2d.detect(
gray_img=frame.gray,
color_img=debug_img,
roi=roi,
)
eye_id = self.g_pool.eye_id
location = result["location"]
result["norm_pos"] = normalize(location, (frame.width, frame.height),
flip_y=True)
result["timestamp"] = frame.timestamp
result["topic"] = f"pupil.{eye_id}.{self.identifier}"
result["id"] = eye_id
result["method"] = "2d c++"
return result
def detect(self, frame):
# convert roi-plugin to detector roi
roi = Roi(*self.g_pool.roi.bounds)
debug_img = frame.bgr if self.g_pool.display_mode == "algorithm" else None
result = self.detector_3d.detect(
gray_img=frame.gray,
timestamp=frame.timestamp,
color_img=debug_img,
roi=roi,
debug=self.is_debug_window_open,
)
eye_id = self.g_pool.eye_id
location = result["location"]
result["norm_pos"] = normalize(location, (frame.width, frame.height),
flip_y=True)
result["topic"] = f"pupil.{eye_id}"
result["id"] = eye_id
result["method"] = "3d c++"
return result
def detect(self, frame, **kwargs):
self._process_focal_length_changes()
# convert roi-plugin to detector roi
roi = Roi(*self.g_pool.roi.bounds)
debug_img = frame.bgr if self.g_pool.display_mode == "algorithm" else None
result = self.detector_3d.detect(
gray_img=frame.gray,
timestamp=frame.timestamp,
color_img=debug_img,
roi=roi,
debug=self.is_debug_window_open,
internal_raw_2d_data=kwargs.get("internal_raw_2d_data", None),
)
# print(kwargs.get("internal_raw_2d_data", None))
#print("-----------------")
#for key, value in result.items():
# #print(key + ": " + str(type(value)))
# if not isinstance(value, dict) and not isinstance(value, bytes):
# print(key + ": " + str(value))
# elif isinstance(value, dict):
# print(key + ":")
# for key2, value2 in value.items():
# print("- " + key2 + ": " + str(value2))
# else:
# print(key + ": " + str(type(value)))
eye_id = self.g_pool.eye_id
location = result["location"]
result["norm_pos"] = normalize(location, (frame.width, frame.height),
flip_y=True)
result["topic"] = f"pupil.{eye_id}.{self.identifier}"
result["id"] = eye_id
result["method"] = "3d c++"
return result
def detect(self, frame, **kwargs):
self._process_focal_length_changes()
# convert roi-plugin to detector roi
roi = Roi(*self.g_pool.roi.bounds)
debug_img = frame.bgr if self.g_pool.display_mode == "algorithm" else None
result = self.detector_3d.detect(
gray_img=frame.gray,
timestamp=frame.timestamp,
color_img=debug_img,
roi=roi,
debug=self.is_debug_window_open,
internal_raw_2d_data=kwargs.get("internal_raw_2d_data", None),
)
eye_id = self.g_pool.eye_id
location = result["location"]
result["norm_pos"] = normalize(location, (frame.width, frame.height),
flip_y=True)
result["topic"] = f"pupil.{eye_id}.{self.identifier}"
result["id"] = eye_id
result["method"] = "3d c++"
return result