def filter_fun(b):
return ((b[2] - b[0]) * (b[3] - b[1])) > 1000
t = 0
saved_time = 0
lock1 = 0
if __name__ == "__main__":
cap = cv2.VideoCapture(0)
#cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1920)
#cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)
detector = MotionDetector(bg_history=20,
brightness_discard_level=25,
bg_subs_scale_percent=0.1,
group_boxes=True,
expansion_step=5)
now = datetime.now()
fourcc = cv2.VideoWriter_fourcc(*'XVID')
output = "backup/backup_" + now.strftime("%d"
"."
"%m"
"."
"%Y"
"_"
"%X") + ".avi"
print(output)
out = cv2.VideoWriter(output, fourcc, 20.0, (640, 480))
# group_boxes=True can be used if one wants to get less boxes, which include all overlapping boxes
@jit(nopython=True)
def filter_fun(b):
return ((b[2] - b[0]) * (b[3] - b[1])) > 1000
if __name__ == "__main__":
cap = cv2.VideoCapture('tmp/helmets-v1-55.mp4')
# cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1920)
# cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)
detector = MotionDetector(bg_history=15,
bg_subs_scale_percent=0.25,
group_boxes=False,
expansion_step=3)
# group_boxes=True can be used if one wants to get less boxes, which include all overlapping boxes
b_height = 512
b_width = 512
res = []
while True:
# Capture frame-by-frame
ret, frame = cap.read()
if frame is None:
break
begin = time()
boxes = detector.detect(frame)
#configPath = "rcnn/mask_rcnn_inception_v2_coco_2018_01_28.pbtxt"
#labels1Path = "rcnn/object_detection_classes_coco.txt"
#labels2Path = "rcnn/object_detection_classes_coco_ru.txt"
weightsPath = "mobilenet/MobileNetSSD_deploy.caffemodel"
configPath = "mobilenet/MobileNetSSD_deploy.prototxt.txt"
labels1Path = "mobilenet/object_detection_classes.txt"
labels2Path = "mobilenet/object_detection_classes.txt"
labels = open(labels1Path).read().strip().split("\n")
labels_ru = open(labels2Path).read().strip().split("\n")
vs = cv2.VideoCapture(0)
(grabbed, frame) = vs.read()
motion_detector = md.MotionDetector(frame)
#object_detector = tod.TensorObjectDetector(weightsPath, configPath)
object_detector = cod.CaffeObjectDetector(weightsPath, configPath)
processor = fp.FrameProcessor(motion_detector, object_detector, labels)
cv2.namedWindow("out_img", cv2.WINDOW_NORMAL)
publisher_worker = publisher.PublisherThread(p.MqttPublisher("1.2.3.4", 1883, "user", "password"),
data_to_publish, publish, 10)
publisher_worker.start()
while True:
(grabbed, img) = vs.read()
#img = cv2.resize(frame, (0, 0), fx=0.8, fy=0.8)
(out_img, out_labels_idxs) = processor.process(img)
@jit(nopython=True)
def filter_fun(b):
return ((b[2] - b[0]) * (b[3] - b[1])) > 300
if __name__ == "__main__":
cap = cv2.VideoCapture('tmp/helmets-v1-55.mp4')
# cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1920)
# cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)
detector = MotionDetector(bg_history=10,
bg_skip_frames=1,
movement_frames_history=2,
brightness_discard_level=5,
bg_subs_scale_percent=0.2,
pixel_compression_ratio=0.1,
group_boxes=True,
expansion_step=5)
# group_boxes=True can be used if one wants to get less boxes, which include all overlapping boxes
b_height = 320
b_width = 320
res = []
fc = dict()
ctr = 0
while True:
# Capture frame-by-frame
ret, frame = cap.read()
import cv2
from imutils.video import VideoStream
from detector import MotionDetector
if __name__ == "__main__":
# cap = cv2.VideoCapture("PATH_TO_VIDEO")
cap = VideoStream(src=0).start()
detector = MotionDetector(bg_history=20,
group_boxes=True,
expansion_step=5,
pixel_compression_ratio=0.4,
min_area=500)
while True:
# Capture frame-by-frame
frame = cap.read()
if frame is None:
break
boxes = detector.detect(frame)
scale = detector.pixel_compression_ratio
for b in boxes:
cv2.rectangle(frame, (int(b[0] / scale), int(b[1] / scale)),
(int(b[0] / scale) + int(b[2] / scale),
int(b[1] / scale) + int(b[3] / scale)), (0, 255, 0),
1)
cv2.imshow('last_frame', frame)
def get_frame_processor(net_config, frame, labels):
motion_detector = md.MotionDetector(frame)
object_detector = cod.CaffeObjectDetector(net_config["model"], net_config["config"])
return fp.FrameProcessor(motion_detector, object_detector, labels)
@jit(nopython=True)
def filter_fun(b):
return ((b[2] - b[0]) * (b[3] - b[1])) > 1000
if __name__ == "__main__":
cap = cv2.VideoCapture('tmp/sample-Elysium.2013.2160p.mkv')
# cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1920)
# cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)
detector = MotionDetector(bg_history=20,
group_boxes=True,
expansion_step=5)
# group_boxes=True can be used if one wants to get less boxes, which include all overlapping boxes
b_height = 512
b_width = 512
res = []
while True:
# Capture frame-by-frame
ret, frame = cap.read()
if frame is None:
break
begin = time()
boxes = detector.detect(frame)