def main():
args = parser()
check_arguments_errors(args)
random.seed(3) # deterministic bbox colors
network, class_names, class_colors = darknet.load_network(
args.config_file,
args.data_file,
args.weights,
batch_size=args.batch_size)
images = load_images(args.input)
index = 0
while True:
# loop asking for new image paths if no list is given
if args.input:
if index >= len(images):
break
image_name = images[index]
else:
image_name = input("Enter Image Path: ")
prev_time = time.time()
image, detections = image_detection(image_name, network, class_names,
class_colors, args.thresh)
if args.save_labels:
save_annotations(image_name, image, detections, class_names)
darknet.print_detections(detections, args.ext_output)
fps = int(1 / (time.time() - prev_time))
print("FPS: {}".format(fps))
if not args.dont_show:
cv2.imshow('Inference', image)
if cv2.waitKey() & 0xFF == ord('q'):
break
index += 1
def inference(darknet_image_queue, detections_queue, fps_queue):
"""
inference the captures into the darknet.
function is also in charge of the printing/writing (fps, caputre time, detections).
"""
# results log
logname = args.export_logname
""" OPTION: """
# each time will open a new txt file
logname_split = args.export_logname.rsplit(".", 1)
index = 0
while 1:
# name_<index>.txt
logname = logname_split[0] + '_' + str(index) + '.' + logname_split[1]
# file not exists
if not os.path.isfile(logname):
break
# trying next index
index += 1
""" OPTION: END """
f = open(logname, "w")
enter_time_queue = [0, 0, 0]
exit_time_queue = [1, 1, 1]
while cap.isOpened():
# get new image from queue
darknet_image = darknet_image_queue.get()
# sample entering time
prev_time = time.time()
enter_time_queue.pop(0)
enter_time_queue.append(prev_time)
# detect image (inference image in neural network)
detections = darknet.detect_image(network,
class_names,
darknet_image,
thresh=args.thresh)
# store result in queue
detections_queue.put(detections)
# calculate fps of passing image
fps = float(1 / (time.time() - prev_time))
exit_time_queue.pop(0)
exit_time_queue.append(time.time())
# store fps in queue
fps_queue.put(int(fps))
# calculate the average fps of 3 last frame (just to follow up)
fps_list = [
1. / (m - n) for m, n in zip(exit_time_queue, enter_time_queue)
]
print("Average FPS over last 3 frames is: {:.2f}".format(
mean(fps_list)))
# store capture time to file (in ms, for ground station)
f.write("time: {}\n".format(str(round(capture_time_queue.get() *
1000))))
# store bbox to file
height_ratio = cap.get(cv2.CAP_PROP_FRAME_HEIGHT) / height
width_ratio = cap.get(cv2.CAP_PROP_FRAME_WIDTH) / width
darknet.print_detections(detections, height_ratio, width_ratio, f)
f.write("\n")
cap.release()
f.close()
print("\nFinished successfully, results: {}".format(logname))
def inference(darknet_image_queue, detections_queue, fps_queue):
while cap.isOpened():
darknet_image = darknet_image_queue.get()
prev_time = time.time()
detections = darknet.detect_image(network, class_names, darknet_image, thresh=args.thresh)
detections_queue.put(detections)
fps = int(1/(time.time() - prev_time))
fps_queue.put(fps)
print("FPS: {}".format(fps))
darknet.print_detections(detections, args.ext_output)
cap.release()
def main():
args = parser()
check_arguments_errors(args)
random.seed(3) # deterministic bbox colors
t = time.time()
network, class_names, class_colors = darknet.load_network(
args.config_file,
args.data_file,
args.weights,
batch_size=args.batch_size)
t1 = time.time()
print("Load weight takes: {}s".format(t1 - t))
image_names = load_images(args.input)
output_dir = args.output_dir
if len(image_names) > 1:
print("Detecting for {} images".format(len(image_names)))
for i in range(len(image_names)):
image_name = image_names[i]
image, detections = image_detection(image_name, network,
class_names, class_colors,
args.thresh)
expression = convert_detections_to_expression(detections)
print(expression)
save_annotations(image_name, image, detections, class_names,
output_dir)
elif len(image_names) == 1:
image_name = image_names[0]
image, detections = image_detection(image_name, network, class_names,
class_colors, args.thresh)
if args.save_labels:
save_annotations(image_name, image, detections, class_names,
output_dir)
darknet.print_detections(detections, args.ext_output)
if len(image_names) < 2 and not args.dont_show:
cv2.imshow('Inference', image)
cv2.waitKey()
expression = convert_detections_to_expression(detections)
expression = normalize_expression(expression)
print(expression)
roots = solve.parse_and_solve_and_round(expression, 0.00001)
print(roots)
latex = convert_infix_to_latex(expression)
print(latex)
t2 = time.time()
print("Detection takes: {}s".format(t2 - t1))
def inference(self):
while self.cap.isOpened():
darknet_image = self.darknet_image_queue.get()
prev_time = time.time()
detections = darknet.detect_image(self.network, self.class_names, self.darknet_image, thresh=self.args.thresh)
self.detections_queue.put(detections)
fps = int(1/(time.time() - prev_time))
self.fps_queue.put(fps)
print("FPS: {}".format(fps))
#印出good,bad,none
darknet.print_detections(detections, self.args.ext_output)
self.cap.release()
def DetectObjects(self):
if self.frame is None:
self.detections = None
return
frameRGB = cv2.cvtColor(self.frame, cv2.COLOR_BGR2RGB)
frameResized = cv2.resize(frameRGB,
(self.darknetWidth, self.darknetHeight),
interpolation=cv2.INTER_LINEAR)
darknet.copy_image_from_bytes(self.darknetImage,
frameResized.tobytes())
self.detections = darknet.detect_image(self.network,
self.classNames,
self.darknetImage,
thresh=self.args.thresh)
darknet.print_detections(self.detections, self.args.ext_output)
def inference(darknet_image_queue, detections_queue, fps_queue):
nth_frame = 0
while cap.isOpened():
darknet_image = darknet_image_queue.get()
prev_time = time.time()
detections = darknet.detect_image(network,
class_names,
darknet_image,
thresh=args.thresh)
original_width = cap.get(cv2.CAP_PROP_FRAME_WIDTH)
original_height = cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
#print("ow, oh", original_width, original_height)
x_scale = original_width / width
y_scale = original_height / height
if len(detections) != 0:
temp_detections = list()
for i in range(len(detections)):
x_detection = detections[i][2][0] * x_scale
y_detection = detections[i][2][1] * y_scale
w_detection = detections[i][2][2] * x_scale
h_detection = detections[i][2][3] * y_scale
temp_detections.append([
detections[i][0], detections[i][1],
(x_detection, y_detection, w_detection, h_detection)
])
detections = temp_detections
if args.output_prefix is not None and len(detections) != 0:
new_file = open(args.output_prefix + str(nth_frame) + ".txt",
mode='w')
for i in range(len(detections)):
x = detections[i][2][0] / cap.get(cv2.CAP_PROP_FRAME_WIDTH)
y = detections[i][2][1] / cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
w = detections[i][2][2] / cap.get(cv2.CAP_PROP_FRAME_WIDTH)
h = detections[i][2][3] / cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
label = detections[i][0]
label = class_names.index(label)
new_file.write("{} {:.4f} {:.4f} {:.4f} {:.4f}\n".format(
label, x, y, w, h))
new_file.close()
detections_queue.put(detections)
fps = int(1 / (time.time() - prev_time))
fps_queue.put(fps)
print("FPS: {}".format(fps))
darknet.print_detections(detections, args.ext_output)
nth_frame += 1
cap.release()
def inference(darknet_image_queue, detections_queue, fps_queue):
while cap.isOpened():
darknet_image = darknet_image_queue.get()
prev_time = time.time()
detections = darknet.detect_image(network, class_names, darknet_image, thresh=args.thresh)
detections_queue.put(detections)
fps = int(1/(time.time() - prev_time))
fps_queue.put(fps)
print("FPS: {}".format(fps))
darknet.print_detections(detections, args.ext_output)
cap.release()
if class_names == "fresh":
fresh = 0
if class_names == "ripe":
ripe = 0
if class_names == "raw":
raw = 0
if class_names == "flowering":
flowering = 0
if class_names == "alternaria":
alternaria = 0
if class_names == "cedar":
cedar = 0
if class_names == "fire-blight":
fire_blight = 0
if class_names == "leaf-roller":
leaf_roller = 0
if class_names == "fungal":
fungal = 0
def main():
pub = rospy.Publisher("/yolo_traffic") #WS_ADDED
args = parser()
check_arguments_errors(args)
random.seed(3) # deterministic bbox colors
network, class_names, class_colors = darknet.load_network(
args.config_file,
args.data_file,
args.weights,
batch_size=args.batch_size)
images = load_images(args.input)
index = 0
while True:
# loop asking for new image paths if no list is given
if args.input:
if index >= len(images):
break
image_name = images[index]
else:
image_name = input("Enter Image Path: ")
prev_time = time.time()
image, detections = image_detection(image_name, network, class_names,
class_colors, args.thresh)
# Call Yolo publishing ftn
rt = publish_yolo(detections) #WS_ADDED
if rt != 0:
pub.publish(rt) #WS_ADDED
else:
ROS_INFO("TRAFFIC LIGHT UNDETECTED")
if args.save_labels:
save_annotations(image_name, image, detections, class_names)
darknet.print_detections(detections, args.ext_output)
fps = int(1 / (time.time() - prev_time))
print("FPS: {}".format(fps))
if not args.dont_show:
cv2.imshow('Inference', image)
if cv2.waitKey() & 0xFF == ord('q'):
break
index += 1
def get_most_confident_bbox(class_name):
# First run through Yolo to check if
image_original, detections = darknet_images.image_detection(
image_name, network, class_names, class_colors, thresh)
darknet.print_detections(detections, ext_output)
# take the detection that had the most confidence and write the coordinates to a txt file
#detections is a list of tuples (class, confidence,(coordinates))
picked_class = -1
for i in range(len(detections)):
_tuple = detections[i]
if _tuple[0] == class_name:
coordinates = _tuple[2]
with open("coordinates.txt", "w+") as coord_file:
# we convert the coordinates to relative because that's how it is used in the C files
coordinates_rel = darknet_images.convert2relative(
image_original, coordinates)
for _coord in coordinates_rel:
coord_file.write(str(_coord) + "\n")
picked_class = i
break
return picked_class, detections
def inference(darknet_image_queue, detections_queue, fps_queue, time_queue,
time_list, pedestrain_frame):
while cap.isOpened():
detect = []
if (not darknet_image_queue.empty()) and (not time_queue.empty()):
darknet_image = darknet_image_queue.get(False)
timestamp = time_queue.get(False)
prev_time = time.time()
if (darknet_image is not None) and (timestamp is not None):
detections = darknet.detect_image(network,
class_names,
darknet_image,
timestamp,
thresh=args.thresh)
"""
for label, confidence, bbox, timestamp in detect_people:
pedestrain_frame.put(timestamp+clip_t*1000)
time_list.append(timestamp)
detect.append((label,confidence,bbox))
detect_people_queue.put(detect, timeout =4)
"""
if len(detections):
for label, confidence, bbox, times in detections:
if pedestrain_frame.empty():
pedestrain_frame.put(times + clip_t * 1000,
block=False)
time_list.append(times)
detect.append((label, confidence, bbox))
if detections_queue.empty():
detections_queue.put(detect, block=False)
fps = int(1 / (time.time() - prev_time))
if fps_queue.empty():
fps_queue.put(fps, block=False)
print("FPS: {}".format(fps))
darknet.print_detections(detections, args.ext_output)
cap.release()
def inference(darknet_image_queue, detections_queue, fps_queue):
while cap.isOpened():
darknet_image = darknet_image_queue.get()
prev_time = time.time()
detections = darknet.detect_image(network,
class_names,
darknet_image,
thresh=args.thresh)
detections_queue.put(detections)
## Code for fire detection
## detections is a list, fire is at position [0][0]
##
if len(detections) > 0: ## IF fire is found in image, trigger alarm
print("\nFIRE FOUND\n")
#os.system("python3 sendText.py")
fps = int(1 / (time.time() - prev_time))
fps_queue.put(fps)
print("FPS: {}".format(fps))
darknet.print_detections(detections, args.ext_output)
darknet.free_image(darknet_image)
cap.release()
if not ret: break
# Calculate for FPS
t_prev = time.time()
# Fix image format
frame_rgb = cv2.cvtColor( frame, cv2.COLOR_BGR2RGB)
frame_resized = cv2.resize( frame_rgb, (width, height))
# convert to darknet format, save to " darknet_image "
darknet_image = darknet.make_image(width, height, 3)
darknet.copy_image_from_bytes(darknet_image, frame_resized.tobytes())
# inference
detections = darknet.detect_image(network, class_names, darknet_image, thresh=thre)
darknet.print_detections(detections, show_coordinates)
darknet.free_image(darknet_image)
# draw bounding box
image = darknet.draw_boxes(detections, frame_resized, class_colors)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
# Show Image and FPS
fps = int(1/(time.time()-t_prev))
cv2.rectangle(image, (5, 5), (75, 25), (0,0,0), -1)
cv2.putText(image, f'FPS {fps}', (10, 20), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2)
cv2.imshow(win_title, cv2.resize(image, show_size))
# press 'q' to quit
if cv2.waitKey(1) == ord('q'):
break
def main():
# read user's input
args = parser()
print("publish to ", args.topic)
# initialize darknet
network, class_names, class_colors = darknet.load_network(args.config_file,
args.data_file,
args.weights,
batch_size=1)
# initialize mqtt client and connect to broker
client = mqtt.Client()
client.connect("localhost")
# read input, webcam or video file
input_path = str2int(args.input)
cap = cv2.VideoCapture(input_path)
# if using webcam, set video fps, width and height
if type(input_path) == int:
cap.set(cv2.CAP_PROP_FPS, 1)
cap.set(3, 416)
cap.set(4, 416)
# get video properties
fps = cap.get(cv2.CAP_PROP_FPS)
w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
# record when to take photo
# take a shot every # seconds
seconds = 3
multiplier = int(fps * seconds)
# set saved video format and properties
if args.out_filename is not None:
video = set_saved_video(cap, args.out_filename, (w, h), 1 / seconds)
if args.out_img is not None:
image_folder = args.out_img + "/" + datetime.datetime.now(
datetime.timezone.utc).astimezone().isoformat(
"T", "seconds").replace(':', '_')
if not os.path.exists(image_folder):
os.makedirs(image_folder)
# Init sql
if args.out_db is not None:
conn = sqlite3.connect(
args.out_db +
datetime.datetime.now(datetime.timezone.utc).astimezone(
).isoformat("T", "seconds").replace(':', '_') + '.sqlite')
cur = conn.cursor()
cur.execute(
'CREATE TABLE inference_result (frameID INTEGER, unique_id TEXT, time TEXT, Walk_stand INTEGER, Car INTEGER, Van INTEGER, Bus INTEGER, Motorcycle INTEGER, Riding_bike INTEGER, Children INTEGER, Skateboarder INTEGER, Queuing INTEGER, Sit INTEGER, Truck INTEGER, Riding_scooter INTEGER)'
)
conn.commit()
count = 0
while cap.isOpened():
count += 1
ret, frame = cap.read()
if not ret:
break
frameID = count
if frameID % multiplier == 0:
if args.vflip:
frame = cv2.flip(frame, 0)
if args.hflip:
frame = cv2.flip(frame, 1)
anno_image, detections = image_detection(frame, network,
class_names, class_colors,
.25)
darknet.print_detections(detections, args.ext_output)
# format result
unique_id = str(uuid.uuid4())
re = count_result(detections, class_names, unique_id)
# send result to mqtt
client.publish(args.topic, json.dumps(re))
# Store result to database
if args.out_db is not None:
# save result to database
cur.execute(
'INSERT INTO inference_result (frameID, unique_id, time, Walk_stand, Car, Van, Bus, Motorcycle, Riding_bike, Children, Skateboarder, Queuing, Sit, Truck, Riding_scooter) values (?,?,?,?,?,?,?,?,?,?,?,?,?,?,?)',
[
frameID, unique_id, re['sampleTime'],
re["count" + class_names[0].title().replace(" ", "")],
re["count" + class_names[1].title().replace(" ", "")],
re["count" + class_names[2].title().replace(" ", "")],
re["count" + class_names[3].title().replace(" ", "")],
re["count" + class_names[4].title().replace(" ", "")],
re["count" + class_names[5].title().replace(" ", "")],
re["count" + class_names[6].title().replace(" ", "")],
re["count" + class_names[7].title().replace(" ", "")],
re["count" + class_names[8].title().replace(" ", "")],
re["count" + class_names[9].title().replace(" ", "")],
re["count" + class_names[10].title().replace(" ", "")],
re["count" + class_names[11].title().replace(" ", "")]
])
conn.commit()
# show image
if not args.dont_show:
cv2.imshow('Inference', frame)
# write video to file
if args.out_filename is not None:
if args.save_infer:
anno_image = cv2.resize(anno_image, (w, h),
interpolation=cv2.INTER_LINEAR)
video.write(anno_image)
else:
video.write(frame)
# write images to folder
if args.out_img is not None:
cv2.imwrite(
image_folder + "/" + str(
round(
datetime.datetime.now(datetime.timezone.utc).
astimezone().timestamp())) + '_' + unique_id +
".jpg", frame)
#cv2.imwrite(image_folder + "/" + unique_id + ".jpg", frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
video.release()
cv2.destroyAllWindows()
index = 0
while True:
# loop asking for new image paths if no list is given
if args.input:
if index >= len(images):
break
image_name = images[index]
else:
image_name = input("Enter Image Path: ")
prev_time = time.time()
image, detections = image_detection(
image_name, network, class_names, class_colors, args.thresh
)
if args.save_labels:
save_annotations(image_name, image, detections, class_names)
darknet.print_detections(detections, args.ext_output)
fps = int(1/(time.time() - prev_time))
print("FPS: {}".format(fps))
if not args.dont_show:
cv2.imshow('Inference', image)
<<<<<<< HEAD
cv2.waitKey(20000)
=======
if cv2.waitKey() & 0xFF == ord('q'):
break
>>>>>>> 1c76abd428ec9957d72b6fe3ec705929f4af4d85
index += 1
if __name__ == "__main__":
# unconmment next line for an example of batch processing
def process(self, ss):
"""
index = 0
while True:
prev_time = time.time()
ss = ImageGrab.grab((0, 0, 1500, 900))
img_np = np.array(ss)
im_rgb = cv2.cvtColor(img_np, cv2.COLOR_BGR2RGB)
cv2.imwrite("temp.jpg", im_rgb)
# loop asking for new image paths if no list is given
image, detections = self.image_detection(
"temp.jpg", self.network, self.class_names, self.class_colors, self.args.thresh
)
if self.args.save_labels:
self.save_annotations(image_name, image, detections, class_names)
darknet.print_detections(detections, self.args.ext_output)
print(detections)
trafficSign = TrafficSign()
if detections != []:
for x in detections:
if x[0] == "ileriSol":
trafficSign.trafficSignArray[2] = 1
cv2.imshow("ileriSol", image[int(x[2][3]):int(x[2][0]), int(x[2][2]):int(x[2][1])])
else:
trafficSign.trafficSignArray[2] = 0
trafficSign.printingAllSigns()
fps = int(1/(time.time() - prev_time))
print("FPS: {}".format(fps))
if not self.args.dont_show:
cv2.imshow('Inference', image)
if cv2.waitKey(25) & 0xFF == ord('q'):
break
index += 1
"""
#ss = ImageGrab.grab((0, 0, 1500, 900))
img_np = np.array(ss)
im_rgb = cv2.cvtColor(img_np, cv2.COLOR_BGR2RGB)
cv2.imwrite("temp.jpg", im_rgb)
# loop asking for new image paths if no list is given
image, detections = self.image_detection("temp.jpg", self.network,
self.class_names,
self.class_colors,
self.args.thresh)
if self.args.save_labels:
self.save_annotations(image_name, image, detections, class_names)
darknet.print_detections(detections, self.args.ext_output)
#print(detections)
#return detections
if detections != []:
for x in detections:
if x[0] == "park":
self.trafficSign.trafficSignArray[0] = 1
else:
self.trafficSign.trafficSignArray[0] = 0
if x[0] == "parkYasak":
self.trafficSign.trafficSignArray[1] = 1
else:
self.trafficSign.trafficSignArray[1] = 0
if x[0] == "durak":
self.trafficSign.trafficSignArray[2] = 1
else:
self.trafficSign.trafficSignArray[2] = 0
if x[0] == "sol":
self.trafficSign.trafficSignArray[3] = 1
else:
self.trafficSign.trafficSignArray[3] = 0
if x[0] == "sag":
self.trafficSign.trafficSignArray[4] = 1
else:
self.trafficSign.trafficSignArray[4] = 0
if x[0] == "solaDonulmez":
self.trafficSign.trafficSignArray[5] = 1
else:
self.trafficSign.trafficSignArray[5] = 0
if x[0] == "sagaDonulmez":
self.trafficSign.trafficSignArray[6] = 1
else:
self.trafficSign.trafficSignArray[6] = 0
if x[0] == "ileriSol":
self.trafficSign.trafficSignArray[7] = 1
else:
self.trafficSign.trafficSignArray[7] = 0
if x[0] == "ileriSag":
self.trafficSign.trafficSignArray[8] = 1
else:
self.trafficSign.trafficSignArray[8] = 0
if x[0] == "dur":
self.trafficSign.trafficSignArray[9] = 1
else:
self.trafficSign.trafficSignArray[9] = 0
if x[0] == "30":
self.trafficSign.trafficSignArray[10] = 1
else:
self.trafficSign.trafficSignArray[10] = 0
if x[0] == "20":
self.trafficSign.trafficSignArray[11] = 1
else:
self.trafficSign.trafficSignArray[11] = 0
if x[0] == "yesil":
self.trafficSign.trafficSignArray[12] = 1
if x[0] == "kirmizi":
self.trafficSign.trafficSignArray[12] = 0
if x[0] == "girisYok":
self.trafficSign.trafficSignArray[13] = 1
else:
self.trafficSign.trafficSignArray[13] = 0
if x[0] == "tasitTrafigineKapali":
self.trafficSign.trafficSignArray[14] = 1
else:
self.trafficSign.trafficSignArray[14] = 0
else:
self.trafficSign.trafficSignArray[0] = 0
self.trafficSign.trafficSignArray[1] = 0
self.trafficSign.trafficSignArray[2] = 0
self.trafficSign.trafficSignArray[3] = 0
self.trafficSign.trafficSignArray[4] = 0
self.trafficSign.trafficSignArray[5] = 0
self.trafficSign.trafficSignArray[6] = 0
self.trafficSign.trafficSignArray[7] = 0
self.trafficSign.trafficSignArray[8] = 0
self.trafficSign.trafficSignArray[9] = 0
self.trafficSign.trafficSignArray[10] = 0
self.trafficSign.trafficSignArray[11] = 0
self.trafficSign.trafficSignArray[12] = -1
self.trafficSign.trafficSignArray[13] = 0
self.trafficSign.trafficSignArray[14] = 0
#cv2.imshow("ileriSol", image[int(x[2][3]):int(x[2][0]), int(x[2][2]):int(x[2][1])])
#cv2.imshow("ileriSol", image[int(x[2][3]):int(x[2][0]), int(x[2][2]):int(x[2][1])])
self.trafficSign.printingAllSigns()
if not self.args.dont_show:
cv2.imshow('Inference', image)