def infer_on_stream(args, client):
"""
Initialize the inference network, stream video to network,
and output stats and video.
. Load Model
. Capture input stream from either camera, video or Image
. Run Async inference per frame.
. Calculate Stats and send image and stats to MQTT or FFMPEG server
Parameters:
args: Command line arguments parsed by `build_argparser()`.
client: connected MQTT client
threshold (float): The minimum threshold for detections.
Returns:
None
"""
# Initialise the class
infer_network = Network()
# Set Probability threshold for detections
prob_threshold = args.prob_threshold
### TODO: Load the model through `infer_network` ###
infer_network.exec_network = infer_network.load_model\
(args.model, args.device, args.cpu_extension)
# extract information about model input layer
(b, c, input_height, input_width) = infer_network.get_input_shape()
### TODO: Handle the input stream ###
# extenstion of input file
input_extension = os.path.splitext(args.input)[1].lower()
supported_vid_exts = ['.mp4', '.mpeg', '.avi', '.mkv']
supported_img_exts = [".bmp",".dib", ".jpeg", ".jp2", ".jpg", ".jpe",\
".png", ".pbm", ".pgm", ".ppm", ".sr", ".ras", ".tiff", ".tif"]
single_image_mode = False
# if input is camera
if args.input.upper() == 'CAM':
capture = cv2.VideoCapture(0)
# if input is video
elif input_extension in supported_vid_exts:
capture = cv2.VideoCapture(args.input)
# if input is image
elif input_extension in supported_img_exts:
single_image_mode = True
capture = cv2.VideoCapture(args.input)
capture.open(args.input)
else:
sys.exit("FATAL ERROR : The format of your input file is not supported" \
"\nsupported extensions are : " + ", ".join(supported_exts))
prev_count = 0
total_persons = 0
### TODO: Loop until stream is over ###
while (capture.isOpened()):
### TODO: Read from the video capture ###
ret, frame = capture.read()
if not ret:
break
### TODO: Pre-process the image as needed ###
image = preprocessing(frame, input_width, input_height)
### TODO: Start asynchronous inference for specified request ###
start_time = time.time()
# run inference
infer_network.exec_net(image)
### TODO: Wait for the result ###
if infer_network.wait() == 0:
infer_time = time.time() - start_time
### TODO: Get the results of the inference request ###
outputs = infer_network.get_output()[0][0]
### Take model output and extract number of detections with confidence exceeding threshold
### and draw bounding boxes around detections
out_image, current_count = apply_threshold(outputs, frame,
prob_threshold)
# show inference time on image
cv2.putText(out_image, "inference time: {:.5f} ms".format(infer_time), (30, 30),\
cv2.FONT_HERSHEY_COMPLEX, 0.5, (0, 255, 0), 1)
### TODO: Extract any desired stats from the results ###
# when any person exit
if current_count < prev_count:
### Topic "person/duration": key of "duration" ###
# send duration to mqtt server client
client.publish("person/duration",
json.dumps({"duration": time.time() - p_start}))
# when new person enters
if current_count > prev_count:
total_persons += current_count - prev_count
p_start = time.time()
prev_count = current_count
### TODO: Calculate and send relevant information on ###
### current_count, total_count and duration to the MQTT server ###
### Topic "person": keys of "count" and "total" ###
client.publish(
"person",
json.dumps({
"count": current_count,
"total": total_persons
}))
### TODO: Send the frame to the FFMPEG server ###
sys.stdout.buffer.write(out_image)
sys.stdout.buffer.flush()
### TODO: Write an output image if `single_image_mode` ###
if single_image_mode:
cv2.imwrite("output_frame.png", out_image)
# release resources
capture.release()
cv2.destroyAllWindows()
client.disconnect()
del infer_network
def infer_on_stream(args, client):
"""
Initialize the inference network, stream video to network,
and output stats and video.
:param args: Command line arguments parsed by `build_argparser()`
:param client: MQTT client
:return: None
"""
global streaming_enabled
# Initialise the class
infer_network = Network()
total_unique_persons = []
use_reidentification = False
# Set Probability threshold for detections
if not args.prob_threshold is None:
prob_threshold = args.prob_threshold
else:
prob_threshold = 0.2
cur_request_id = 0
last_detection_time = None
duration = 0
start = None
single_image_mode = False
show_info = False
if args.show_info:
show_info = args.show_info
message = None
if args.message:
message = args.message
if args.input == 'CAM':
input_stream = 0
elif args.input.endswith('.jpg') or args.input.endswith(
'.bmp') or args.input.endswith('.png'):
single_image_mode = True
input_stream = args.input
# Checks for video file
else:
input_stream = args.input
assert os.path.isfile(args.input), "Specified input file doesn't exist"
### TODO: Load the model through `infer_network` ###
n, c, h, w = infer_network.load_model(args.model, args.device, 1, 1,
cur_request_id,
args.cpu_extension)[1]
# Intialize class for reidentification
networkReIdentification = None
identification_input_shape = None
if args.reident_model:
networkReIdentification = Network()
networkReIdentification.load_model(args.reident_model, args.device, 1,
1, cur_request_id,
args.cpu_extension)
identification_input_shape = networkReIdentification.get_input_shape()
use_reidentification = True
### TODO: Handle the input stream ###
if not single_image_mode:
cap = cv2.VideoCapture(input_stream)
if input_stream:
cap.open(args.input)
if not cap.isOpened():
log.error("ERROR! Unable to open video source")
detection_frame_count = 0
total_frame_count = 0
previous_detection_time = None
last_person_counts = []
average_person_count = 0
detection_time = None
total_seconds_elapsed_for_detection = 0
# Parameters for duration
max_len = 40
track_threshold = 0.2
track = deque(maxlen=max_len)
### TODO: Loop until stream is over ###
while cap.isOpened():
### TODO: Read from the video capture ###
flag, frame = cap.read()
if not flag:
break
### TODO: Pre-process the image as needed ###
image = preprocessing(frame, h, w)
### TODO: Start asynchronous inference for specified request ###
inf_start = time.time()
infer_network.exec_network(cur_request_id, image)
### TODO: Wait for the result ###
output_img = frame
if infer_network.wait(cur_request_id) == 0:
### TODO: Get the results of the inference request ###
det_time = time.time() - inf_start
result = infer_network.get_output(cur_request_id)
### TODO: Extract any desired stats from the results ###
image_h, image_w, _ = frame.shape
num_detections = 0
for box in result[0][0]:
label = box[1]
conf = box[2]
if label == 1:
if (conf > prob_threshold):
x_min = int(box[3] * image_w)
y_min = int(box[4] * image_h)
x_max = int(box[5] * image_w)
y_max = int(box[6] * image_h)
dist = (y_max - y_min) / (y_min + y_max)
color = (0, dist * 255, 255 - dist * 255)
if use_reidentification:
try:
if conf > 0.85:
crop_person = frame[y_min:y_max,
x_min:x_max]
total_unique_persons = reidentification(
cur_request_id,
networkReIdentification,
crop_person,
identification_input_shape,
total_unique_persons, conf)
except Exception as err:
pass
cv2.rectangle(frame, (x_min, y_min),
(x_max, y_max), color, int(1))
num_detections += 1
last_detection_time = datetime.now()
if start is None:
start = time.time()
time.clock()
else:
label_box_pos = None
if last_detection_time is not None:
second_diff = (datetime.now() -
last_detection_time).total_seconds()
if second_diff >= 1.5:
if start is not None and num_detections == 0:
elapsed = time.time() - start
client.publish(
"person/duration",
json.dumps({"duration": elapsed}))
start = None
last_detection_time = None
person_counts = num_detections
overlay = output_img.copy()
if show_info:
cv2.putText(overlay, message, (10, 40), FONT, 1,
(250, 250, 250), 2, cv2.LINE_AA)
cv2.putText(overlay,
'Person[s] found: {}'.format(person_counts),
(10, overlay.shape[0] - 40), FONT, 1,
(255, 255, 255), 1, cv2.LINE_AA)
cv2.putText(
overlay,
str(datetime.now().strftime(
"%A, %d. %B %Y %I:%M:%S %p")),
(10, overlay.shape[0] - 20), FONT, 1, (250, 250, 250),
1, cv2.LINE_AA)
cv2.addWeighted(overlay, ALPHA, output_img, 1 - ALPHA, 0,
output_img)
if len(last_person_counts) > 10:
last_person_counts = last_person_counts[
1:len(last_person_counts) - 1]
last_person_counts.append(person_counts)
average_person_count = int(
sum(last_person_counts) / len(last_person_counts))
client.publish(
"person",
json.dumps({
"count": str(person_counts),
"total": len(total_unique_persons)
}))
### TODO: Send the frame to the FFMPEG server ###
if streaming_enabled:
sys.stdout.buffer.write(output_img)
sys.stdout.flush()
pass
if cap:
cap.release()
cv2.destroyAllWindows()
client.disconnect()
infer_network.dispose()
### TODO: Write an output image if `single_image_mode` ###
elif single_image_mode:
frame = cv2.imread(input_stream)
image = preprocessing(frame, h, w)
infer_network.exec_network(0, image)
if infer_network.wait(0) == 0:
result = infer_network.get_output(0)
output_img, person_counts = get_draw_boxes_on_image(
result, frame, prob_threshold, True)
cv2.imwrite('output_image.jpg', output_img)
def infer_on_stream(args, client):
"""
Initialize the inference network, stream video to network,
and output stats and video.
:param args: Command line arguments parsed by `build_argparser()`
:param client: MQTT client
:return: None
"""
# Initialise the class
infer_network = Network()
# Set Probability threshold for detections
#prob_threshold = args.prob_threshold
single_image = False
### TODO: Load the model through `infer_network` ###
infer_network.load_model(args.model, args.device, CPU_EXTENSION)
### TODO: Handle the input stream ###
net_input_shape = infer_network.get_input_shape()
#Check for CAM, image or video
if args.input == 'CAM':
input_stream = 0
elif args.input.endswith('.jpg') or args.input.endswith('.bmp'):
single_image = True
input_stream = args.input
else:
input_stream = args.input
if not os.path.isfile(args.input):
log.error("Specified input file doesn't exist")
sys.exit(1)
cap = cv2.VideoCapture(input_stream)
if input_stream:
cap.open(args.input)
if not cap.isOpened():
log.error("Unable to open source")
width = int(cap.get(3))
height = int(cap.get(4))
out = cv2.VideoWriter('out.mp4', 0x00000021, 10, (width, height))
global incident_flag, quantity, timesnap, timer, ticks, pt
pt = args.prob_threshold
incident_flag = False
quantity = 0
total = 0
timesnap = 0
timer = False
ticks = 0
curr_count = 0
doneCounter = False
start_time = 0
### TODO: Loop until stream is over ###
while cap.isOpened():
### TODO: Read from the video capture ###
timesnap += 1
flag, original_frame = cap.read()
if not flag:
break
if cv2.waitKey(25) & 0xFF == ord('q'):
break
### TODO: Pre-process the image as needed ###
frame = cv2.resize(original_frame,
(net_input_shape[3], net_input_shape[2]))
frame = frame.transpose((2, 0, 1))
frame = frame.reshape(1, *frame.shape)
### TODO: Start asynchronous inference for specified request ###
infer_network.exec_network(frame)
### TODO: Wait for the result ###
inf_start = time.time()
if infer_network.wait() == 0:
det_time = time.time() - inf_start
### TODO: Get the results of the inference request ###
result = infer_network.get_output()
### TODO: Extract any desired stats from the results ###
out_frame = draw_boundingBox(result, original_frame, height, width)
inf_time_message = "Inference time: {:.3f}ms".format(det_time *
1000)
cv2.putText(out_frame, inf_time_message, (15, 15),
cv2.FONT_HERSHEY_COMPLEX, 0.5, (200, 10, 10), 1)
### TODO: Calculate and send relevant information on ###
out.write(out_frame)
cv2.imshow('frame', out_frame)
curr_count = detect_person(result, curr_count)
if incident_flag and not doneCounter:
start_time = time.time()
total += 1
print("Total: {}".format(total))
doneCounter = True
json.dumps({"total": total})
#client.publish("person",json.dumps({"total":total}))
if not incident_flag and doneCounter and total >= 1:
doneCounter = False
duration = int(time.time() - start_time)
print("Count: {}".format(curr_count))
# Publish messages to the MQTT server
#client.publish("person/duration",
#json.dumps({"duration": duration}))
# client.publish("person",json.dumps({"count":curr_count}))
### current_count, total_count and duration to the MQTT server ###
### Topic "person": keys of "count" and "total" ###
### Topic "person/duration": key of "duration" ###
#sys.stdout.buffer.write(out_frame)
#sys.stdout.flush()
### TODO: Send the frame to the FFMPEG server ###
### TODO: Write an output image if `single_image_mode` ###
if single_image:
cv2.imwrite('out_image.jpg', out_frame)
out.release()
cap.release()
cv2.destroyAllWindows()
