def __init__(self, detection_pipeline, classes_images_path,
classes_sound_path):
self.__detection_pipeline = detection_pipeline
self.__plotter = Plotter(num_classes=20, bgr=True)
self.__logger = Logger()
self.__cfd = CrossFrameDetector(num_classes=20,
frame_history_count=5,
frames_threshold=2)
self.__classes_sound_path = classes_sound_path
self.__classes_images = self.__load_classes_images(classes_images_path)
self.__notification_queue = []
def test_plot_detections_top_side(self):
plt = Plotter(num_classes=20)
test_detections = [
{
'coordinates': [0, 0, 250, 400],
'confidence': 0.998,
'class_id': 16,
'label': 'contramano'
},
]
test_image_path = 'tests/data/test.jpg'
image = cv2.imread(test_image_path)
drawn_image = plt.plot_detections(image, test_detections)
assert drawn_image.shape == image.shape
def test_plot_detections_bgr(self):
plt = Plotter(num_classes=20, bgr=True)
test_detections = [{
'coordinates': [200, 100, 300, 200],
'confidence': 0.998,
'class_id': 16,
'label': 'contramano'
}, {
'coordinates': [300, 800, 400, 950],
'confidence': 0.783,
'class_id': 13,
'label': 'cocoa'
}]
test_image_path = 'tests/data/test.jpg'
image = cv2.imread(test_image_path)
drawn_image = plt.plot_detections(image, test_detections)
assert drawn_image.shape == image.shape
def test_save_plot_detections(self):
plt = Plotter(num_classes=20)
test_detections = [{
'coordinates': [200, 100, 300, 200],
'confidence': 0.998,
'class_id': 16,
'label': 'contramano'
}, {
'coordinates': [300, 800, 400, 950],
'confidence': 0.783,
'class_id': 13,
'label': 'cocoa'
}]
test_image_path = 'tests/data/test.jpg'
image = cv2.imread(test_image_path)
output_filename = './tests/data/test_output.jpg'
drawn_image = plt.plot_detections(image,
test_detections,
output=output_filename)
assert os.path.exists(output_filename) == True
def test_plot_detections_right_side(self):
plt = Plotter(num_classes=20)
test_detections = [{
'coordinates': [450, 400, 500, 500],
'confidence':
0.998,
'class_id':
16,
'label':
'test-long-long-long-long-long-long-long-long-long-long-label'
}, {
'coordinates': [300, 800, 400, 950],
'confidence': 0.783,
'class_id': 13,
'label': 'cocoa'
}]
test_image_path = 'tests/data/test.jpg'
image = cv2.imread(test_image_path)
drawn_image = plt.plot_detections(image, test_detections)
assert drawn_image.shape == image.shape
class Detector:
def __init__(self, detection_pipeline, classes_images_path,
classes_sound_path):
self.__detection_pipeline = detection_pipeline
self.__plotter = Plotter(num_classes=20, bgr=True)
self.__logger = Logger()
self.__cfd = CrossFrameDetector(num_classes=20,
frame_history_count=5,
frames_threshold=2)
self.__classes_sound_path = classes_sound_path
self.__classes_images = self.__load_classes_images(classes_images_path)
self.__notification_queue = []
def detect_image(self,
image,
output=None,
show_confidence=False,
return_image=False):
detections = self.__detection_pipeline.detect_objects_in_image(image)
if output:
filename, extension = os.path.splitext(output)
if extension == '.json':
self.__logger.save_detections_to_json(detections, output)
elif extension == '.txt':
h, w, _ = image.shape
self.__logger.save_detections_to_yolo(detections, (w, h),
output)
elif extension == '.csv':
self.__logger.save_detections_to_csv(detections, output)
elif extension == '.jpg':
detections_image = self.__plotter.plot_detections(
image, detections, draw_confidence=show_confidence)
cv2.imwrite(output, detections_image)
else:
raise FormatNotSupportedException(
'Output extension must be .json .txt .csv or .jpg')
if return_image:
detections_image = self.__plotter.plot_detections(
image, detections, draw_confidence=show_confidence)
return detections_image, detections
return detections
def detect_video_feed(self,
video_feed,
show_output=False,
sound_notifications=False,
output=None,
output_csv=None,
show_confidence=False,
skip_frames=False):
cap = cv2.VideoCapture(video_feed)
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
fps = int(cap.get(cv2.CAP_PROP_FPS))
if skip_frames:
cap.release()
if output:
fourcc = cv2.VideoWriter_fourcc(*'XVID')
out = cv2.VideoWriter(output, fourcc, fps, (width, height))
if output_csv:
csv_file = open(output_csv, 'w')
titles = ['frame_id', 'class_id', 'x1', 'y1', 'x2', 'y2']
csv_file.write(', '.join(titles) + '\n')
if show_output:
# from screeninfo import get_monitors
# monitor = get_monitors()[0]
# screen_w, screen_h, = monitor.width, monitor.height
screen_w, screen_h, = 1920, 1080
window_name = 'Detector'
cv2.namedWindow(window_name, cv2.WND_PROP_FULLSCREEN)
cv2.moveWindow(window_name, screen_w - 1, screen_h - 1)
cv2.setWindowProperty(window_name, cv2.WND_PROP_FULLSCREEN,
cv2.WINDOW_FULLSCREEN)
if sound_notifications:
sound_thread = threading.Thread(
target=self.__notification_thread, args=())
sound_thread.daemon = True
self.sound_notifications_lock = True
sound_thread.start()
frame_id = 0
while (True):
# Capture frame-by-frame
if skip_frames:
cap = cv2.VideoCapture(video_feed)
ret, frame = cap.read()
frame_id += 1
if skip_frames:
cap.release()
if not ret:
break
print('\r>> Processing frame Nº {}'.format(frame_id), end='')
detections = self.__detection_pipeline.detect_objects_in_image(
frame)
if output_csv:
for detection in detections:
class_id = detection['class_id']
coordinates = detection['coordinates']
line = ', '.join(
[str(x) for x in [frame_id, class_id, *coordinates]])
csv_file.write(line + '\n')
self.__cfd.register_detections(detections)
frame = self.__plotter.plot_detections(
frame, detections, draw_confidence=show_confidence)
all_cfd_detections, current_cfd_detections = self.__cfd.get_detections(
)
if sound_notifications:
for detection_id in current_cfd_detections:
self.__notification_queue.append(detection_id)
cfd_frame = self.__draw_cfd_detections(frame, all_cfd_detections)
# Display the resulting frame
if show_output:
cv2.imshow(window_name, frame)
if output:
out.write(frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
if output_csv:
csv_file.close()
if output:
out.release()
cap.release()
cv2.destroyAllWindows()
cv2.waitKey(1)
cv2.waitKey(1)
cv2.waitKey(1)
cv2.waitKey(1)
self.__cfd.reset()
if sound_notifications:
self.sound_notifications_lock = False
return True
def __notification_thread(self):
while self.sound_notifications_lock:
if len(self.__notification_queue) > 0:
class_id = self.__notification_queue.pop()
notification_path = os.path.join(self.__classes_sound_path,
'notification.mp3')
notification_sound = AudioSegment.from_mp3(notification_path)
play(notification_sound)
class_sound_path = os.path.join(self.__classes_sound_path,
str(class_id) + '.mp3')
class_name_sound = AudioSegment.from_mp3(class_sound_path)
play(class_name_sound)
def __load_classes_images(self, path):
classes_images = {}
images_list = glob.glob(os.path.join(path, '*.jpg'))
for image_path in images_list:
class_id = int(os.path.basename(image_path)[:-4])
image = cv2.imread(image_path)
image = cv2.copyMakeBorder(image,
10,
10,
10,
10,
cv2.BORDER_CONSTANT,
value=[0, 0, 0])
classes_images[class_id] = image
return classes_images
def __draw_cfd_detections(self, frame, cfd_detections):
cfd_detection_count = len(cfd_detections)
if cfd_detection_count > 0:
frame_height, frame_width, _ = frame.shape
class_image_size = int(frame_height / 4)
aux = np.zeros(shape=(class_image_size,
class_image_size * cfd_detection_count, 3),
dtype='uint8')
start_h, start_w = 0, 0
for class_id in cfd_detections:
img = self.__classes_images[class_id]
img = cv2.resize(img, (class_image_size, class_image_size))
end_w = start_w + img.shape[1]
aux[start_h:img.shape[0], start_w:end_w] = img
start_w = start_w + img.shape[1]
frame[frame_height - class_image_size:, :class_image_size *
cfd_detection_count] = aux
return frame