def tracking():
"""Perform object tracking"""
print(LICENSE_MSG)
args = parse_args()
try:
write_mode = 'wb' if args.force else 'xb'
with open(args.output_file, write_mode) as output_file:
video = open_video(args.video_file)
read_frame(video)
print(SET_SCALE_MSG)
press_enter()
print('<scale window displayed>')
distance = get_scale_distance()
print('\nThe line drawn has a distance of {:.1f} pixels.'.format(
distance))
measure = float(
input('Tell me how many metres this should represent. > '))
scale = distance / measure
print(ROI_BOX_MSG)
press_enter()
print('<object tracking window displayed>')
bbox = select_bounding_box()
if args.algorithm == 'KCF':
tracker = cv2.TrackerKCF_create()
elif args.algorithm == 'MIL':
tracker = cv2.TrackerMIL_create()
elif args.algorithm == 'Median-Flow':
tracker = cv2.TrackerMedianFlow_create()
else:
raise ValueError('Unknown algorithm type')
if args.suppress_live_plot:
print(TRACKING_MSG)
else:
print(TRACKING_MSG_W_PLOT)
press_enter()
print('<object tracking window displayed>')
speed_up = 1
if args.speed_up:
speed_up = int(args.speed_up[:-1])
points = track_video(video, tracker, bbox, scale,
args.suppress_live_plot, args.algorithm,
speed_up)
np.save(output_file, np.asarray(points))
print(LAST_PLOT_MSG)
press_enter()
fig = plt.figure()
axes = fig.gca(projection='3d')
axes.plot(points.T[1], points.T[2], zs=points.T[0])
axes.set_title('Tracked object motion')
axes.set_aspect('equal')
xmin, xmax = axes.get_xlim()
x_range = abs(xmin - xmax)
ymin, ymax = axes.get_ylim()
y_range = abs(ymin - ymax)
half_diff = abs(x_range - y_range) / 2
if x_range > y_range:
axes.set_ylim(ymin - half_diff, ymax + half_diff)
if y_range > x_range:
axes.set_xlim(xmin - half_diff, xmax + half_diff)
plt.show()
except FileExistsError:
print('This directory already contains a file named: {}'.format(
args.output_file))
print('Please move, rename, or delete this file and try again.')
#Initialize the looping variable to one.
loop = 1
#Socket Setup - socket was used because the client script needs libraries of opencv that were not installed.
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
host = socket.gethostname()
port = 9547
s.bind(('', port))
s.listen(5)
#Wait until a connection request is received, then proceed.
clientSocket, addr = s.accept()
print("got a connection from %s" % str(addr))
tracker = cv2.TrackerKCF_create() #create tracker object
cap = cv2.VideoCapture(1) #load video (0 or 1 for camera device number)
ok, frame = cap.read() #read one frame of the stream
listofbbox = [] #initialize the vector of bbox lists
#Setup code so that the output can be saved to file.
fourcc = cv2.VideoWriter_fourcc(*'XVID')
out = cv2.VideoWriter('output12_10.avi', fourcc, 20.0, (640, 480))
#Select the bbox for each character in order. Press enter after each one.
for i in range(0, int(chars)):
listofbbox.append(cv2.selectROI(
frame, False)) #select location of ROI (region of interest)
#Set bbox equal to the first bbox specified.
bbox = listofbbox[0]
tracker_types = [
'BOOSTING', 'MIL', 'KCF', 'TLD', 'MEDIANFLOW', 'GOTURN', 'MOSSE', 'CSRT'
]
tracker_type = tracker_types[7]
if int(minor_ver) < 3:
tracker = cv2.TrackerMOSSE_create()
#tracker = cv2.Tracker_create(tracker_type)
else:
if tracker_type == 'BOOSTING':
tracker = cv2.TrackerBoosting_create()
elif tracker_type == 'MIL':
tracker = cv2.TrackerMIL_create()
elif tracker_type == 'KCF':
tracker = cv2.TrackerKCF_create()
elif tracker_type == 'TLD':
tracker = cv2.TrackerTLD_create()
elif tracker_type == 'MEDIANFLOW':
tracker = cv2.TrackerMedianFlow_create()
elif tracker_type == 'GOTURN':
tracker = cv2.TrackerGOTURN_create()
elif tracker_type == 'MOSSE':
tracker = cv2.TrackerMOSSE_create()
elif tracker_type == 'CSRT':
tracker = cv2.TrackerCSRT_create()
print("type is ", tracker_type)
video = cv2.VideoCapture(1)
cv2.namedWindow('result')
cv2.setMouseCallback('result', draw_rectangle)
if not video.isOpened():
def run(self):
# Capture, process, display loop
while True:
# Read a new frame
ok, self.frame = self.video.read()
display = self.frame.copy(
) # Frame we'll do all the graphical drawing to
data_display = np.zeros_like(
display, dtype=np.uint8) # Black screen to display data
if not ok:
break
# Start FPS timer
timer = cv2.getTickCount()
if self.bg is None:
cv2.putText(display,
"Press 'r' to reset the foreground extraction.",
self.positions['hand_pose'],
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 127, 64), 2)
cv2.imshow("display", display)
k = cv2.waitKey(1) & 0xff
if k == 27: break # ESC pressed
elif k == 114 or k == 108:
# r pressed
self.bg = self.frame.copy()
self.hand_bbox = (116, 116, 170, 170)
self.is_tracking = False
else:
# Extract the foreground
foreground, mask = self.extract_foreground()
foreground_display = foreground.copy()
# Get hand from mask using the bounding box
hand_crop = mask[int(self.hand_bbox[1]):int(self.hand_bbox[1] +
self.hand_bbox[3]),
int(self.hand_bbox[0]):int(self.hand_bbox[0] +
self.hand_bbox[2])]
# Update tracker
if self.is_tracking:
tracking, self.hand_bbox = self.tracker.update(foreground)
try:
# Resize cropped hand and make prediction on gesture
hand_crop_resized = np.expand_dims(cv2.resize(
hand_crop, (54, 54)),
axis=0).reshape(
(1, 54, 54, 1))
prediction = self.recognizer.predict(hand_crop_resized)
predi = prediction[0].argmax(
) # Get the index of the greatest confidence
gesture = self.CLASSES[predi]
for i, pred in enumerate(prediction[0]):
# Draw confidence bar for each gesture
barx = self.positions['hand_pose'][0]
bary = 60 + i * 60
bar_height = 20
bar_length = int(
400 *
pred) + barx # calculate length of confidence bar
# Make the most confidence prediction green
if i == predi:
colour = (0, 255, 0)
else:
colour = (0, 0, 255)
cv2.putText(
data_display,
"{}: {}".format(self.CLASSES[i], pred),
(self.positions['hand_pose'][0], 30 + i * 60),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (255, 255, 255), 2)
cv2.rectangle(data_display, (barx, bary),
(bar_length, bary - bar_height), colour,
-1, 1)
cv2.putText(display, "hand pose: {}".format(gesture),
self.positions['hand_pose'],
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)
cv2.putText(foreground_display,
"hand pose: {}".format(gesture),
self.positions['hand_pose'],
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)
except Exception as ex:
cv2.putText(display, "hand pose: error",
self.positions['hand_pose'],
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)
cv2.putText(foreground_display, "hand pose: error",
self.positions['hand_pose'],
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)
# Draw bounding box
p1 = (int(self.hand_bbox[0]), int(self.hand_bbox[1]))
p2 = (int(self.hand_bbox[0] + self.hand_bbox[2]),
int(self.hand_bbox[1] + self.hand_bbox[3]))
cv2.rectangle(foreground_display, p1, p2, (255, 0, 0), 2, 1)
cv2.rectangle(display, p1, p2, (255, 0, 0), 2, 1)
# Calculate difference in hand position
hand_pos = ((p1[0] + p2[0]) // 2, (p1[1] + p2[1]) // 2)
mouse_change = ((p1[0] + p2[0]) // 2 -
self.positions['null_pos'][0],
self.positions['null_pos'][0] -
(p1[1] + p2[1]) // 2)
# Draw hand moved difference
cv2.circle(display, self.positions['null_pos'], 5, (0, 0, 255),
-1)
cv2.circle(display, hand_pos, 5, (0, 255, 0), -1)
cv2.line(display, self.positions['null_pos'], hand_pos,
(255, 0, 0), 5)
# Calculate Frames per second (FPS)
fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer)
# Display FPS on frame
cv2.putText(foreground_display, "FPS : " + str(int(fps)),
self.positions['fps'], cv2.FONT_HERSHEY_SIMPLEX,
0.65, (50, 170, 50), 2)
cv2.putText(display, "FPS : " + str(int(fps)),
self.positions['fps'], cv2.FONT_HERSHEY_SIMPLEX,
0.65, (50, 170, 50), 2)
# Display pause command text
cv2.putText(
foreground_display,
"hold 'r' to recalibrate until the screen is black",
(15, 400), cv2.FONT_HERSHEY_SIMPLEX, 0.65, (0, 0, 255), 2)
cv2.putText(foreground_display, "to recalibrate", (15, 420),
cv2.FONT_HERSHEY_SIMPLEX, 0.65, (0, 0, 255), 2)
cv2.putText(foreground_display,
"press 'p' to return to paused state", (15, 450),
cv2.FONT_HERSHEY_SIMPLEX, 0.65, (0, 0, 255), 2)
# Display foreground_display
cv2.imshow("foreground_display", foreground_display)
cv2.imshow("display", display)
# Display result
cv2.imshow("data", data_display)
k = cv2.waitKey(1) & 0xff
if k == 27: break # ESC pressed
elif k == 114 or k == 108: # r pressed
self.bg = self.frame.copy()
self.hand_bbox = (116, 116, 170, 170)
self.is_tracking = False
elif k == 116: # t pressed
# Initialize tracker with first frame and bounding box
self.tracker = cv2.TrackerKCF_create()
self.tracker.init(self.frame, self.hand_bbox)
elif k == 112: # p pressed
# Reset to paused state
self.is_tracking = False
self.bg = None
cv2.destroyAllWindows()
elif k != 255:
print(k)
def Analise(result, ui=False, vid_folder="/content/TRAIN_0"):
data = {
'ATA': 0.0,
'F': 0.0,
'F1': 0.0,
'OTP': 0.0,
'OTA': 0.0,
'Deviation': 0.0,
'PBM': 0.0,
'FPS': 0.0,
'Ms': 0,
'fp': 0,
'tp': 0,
'fn': 0,
'g': 0
}
result[1] = data
tracker_type = result[0]
result = result[1:]
seq_ID = result[1]["my_object"]["video"]
del result[1]["my_object"]["video"]
frames_folder = os.path.join(vid_folder, "frames", seq_ID)
anno_file = os.path.join(vid_folder, "anno", seq_ID + ".txt")
anno = np.loadtxt(anno_file, delimiter=",")
frames_list = [
frame for frame in os.listdir(frames_folder) if frame.endswith(".jpg")
]
if not len(anno) == len(frames_list):
print("Not the same number of frames and annotation!")
return
# Define an initial bounding box
bbox = (anno[0][0], anno[0][1], anno[0][2], anno[0][3])
frame = cv2.imread(os.path.join(frames_folder, "0.jpg"))
if tracker_type == "MOSSE":
tracker = cv2.TrackerMOSSE_create()
elif tracker_type == "TLD":
tracker = cv2.TrackerTLD_create()
elif tracker_type == "GOTURN":
tracker = cv2.TrackerGOTURN_create()
elif tracker_type == "BOOSTING":
tracker = cv2.TrackerBoosting_create()
elif tracker_type == "MIL":
tracker = cv2.TrackerMIL_create()
elif tracker_type == "KCF":
tracker = cv2.TrackerKCF_create()
elif tracker_type == "MEDIANFLOW":
tracker = cv2.TrackerMedianFlow_create()
elif tracker_type == "CSRT":
tracker = cv2.TrackerCSRT_create()
file_path = rndStr() + ".json"
file1 = open(file_path, 'w')
file1.writelines(json.dumps(result[1]))
file1.close()
fs = cv2.FileStorage(file_path, cv2.FILE_STORAGE_READ)
tracker.read(fs.getFirstTopLevelNode())
os.remove(file_path)
ok = tracker.init(frame, bbox)
if not ok:
print("Initialisation error")
continue
data["Ms"] += 1
data["tp"] += 1
data["ATA"] += IOU(bbox, bbox)
data["Deviation"] += NCDist(bbox, bbox)
data["F1"] += F1(bbox, bbox)
data["PBM"] += 1 - L1(bbox, bbox)
for i in range(1, len(frames_list)):
frame_file = str(i) + ".jpg"
imgs_file = os.path.join(frames_folder, frame_file)
frame = cv2.imread(imgs_file)
anno_bbox = (anno[i][0], anno[i][1], anno[i][2], anno[i][3])
x_min = int(anno[i][0])
x_max = int(anno[i][2] + anno[i][0])
y_min = int(anno[i][1])
y_max = int(anno[i][3] + anno[i][1])
cv2.rectangle(frame, (x_min, y_min), (x_max, y_max), (0, 0, 255), 2, 1)
# Start timer
timer = cv2.getTickCount()
# Update tracker
ok, bbox = tracker.update(frame)
# Calculate Frames per second (FPS)
fpS = cv2.getTickFrequency() / (cv2.getTickCount() - timer)
# Draw bounding box
if ok:
# Tracking success
p1 = (int(bbox[0]), int(bbox[1]))
p2 = (int(bbox[0] + bbox[2]), int(bbox[1] + bbox[3]))
if ui:
cv2.rectangle(frame, p1, p2, (255, 0, 0), 2, 1)
else:
# Tracking failure
# print("Tracking failure detected")
if True:
data["fn"] += 1
if ui:
cv2.putText(frame, "Tracking failure detected", (100, 80),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)
# Display tracker type on frame
if ui:
cv2.putText(frame, tracker_type + " Tracker", (100, 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2)
# Display FPS on frame
cv2.putText(frame, "FPS : " + str(int(fpS)), (100, 50),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2)
# Display result
cv2.imshow("Tracking", frame)
iou = IOU(bbox, anno_bbox)
data["ATA"] += iou
data["FPS"] += fpS
data["F1"] += F1(bbox, anno_bbox)
data["PBM"] += 1 - (L1(bbox, anno_bbox) if iou > 0 else Th(
bbox, anno_bbox)) / Th(bbox, anno_bbox)
if True:
data["g"] += 1
if iou >= 0.5:
data["tp"] += 1
elif iou > 0:
data["OTP"] += iou
data["Deviation"] += NCDist(bbox, anno_bbox)
data["Ms"] += 1
else:
data["fp"] += 1
# Exit if ESC pressed
if ui:
k = cv2.waitKey(1) & 0xff
if k == 27:
sys.exit()
# print(tracker_type, i)
data["F"] = F(data["tp"][j], data["fp"][j], data["fn"][j])
data["F1"] /= len(frames_list)
data["OTA"] = 1 - (data["fp"][j] + data["fn"][j]) / data["g"][j]
data["OTP"] /= data["Ms"][j]
data["ATA"] /= len(frames_list)
data["FPS"] /= len(frames_list)
data["Deviation"] = 1 - data["Deviation"][j] / data["Ms"][j]
if data["Deviation"] < -1:
data["Deviation"] = -1
data["PBM"] /= len(frames_list)
result[0].update(result[1]["my_object"])
result = result[0]
def start(self):
(major_ver, minor_ver, subminor_ver) = (cv2.__version__).split('.')
# Задаем трекер.
tracker_types = [
'BOOSTING', 'MIL', 'KCF', 'TLD', 'MEDIANFLOW', 'GOTURN'
]
tracker_type = tracker_types[4]
if int(minor_ver) < 3:
tracker = cv2.Tracker_create(tracker_type)
else:
if tracker_type == 'BOOSTING':
tracker = cv2.TrackerBoosting_create()
if tracker_type == 'MIL':
tracker = cv2.TrackerMIL_create()
if tracker_type == 'KCF':
tracker = cv2.TrackerKCF_create()
if tracker_type == 'TLD':
tracker = cv2.TrackerTLD_create()
if tracker_type == 'MEDIANFLOW':
tracker = cv2.TrackerMedianFlow_create()
if tracker_type == 'GOTURN':
tracker = cv2.TrackerGOTURN_create()
# Захват камеры
video = cv2.VideoCapture(0)
if not video.isOpened():
print("Камера не в 0 порту")
sys.exit()
# берем первый кадр
ok, frame = video.read()
if not ok:
print('Ошибка чтения видео')
sys.exit()
# Рамкой выделяем объект
bbox = (287, 23, 86, 320)
bbox = cv2.selectROI(frame, False)
# Инициализируем трекер с помощью первого кадра и ограничительной рамки
ok = tracker.init(frame, bbox)
while True:
# Читаем новый фрейм
ok, frame = video.read()
if not ok:
break
# Запуск таймера
timer = cv2.getTickCount()
# Обновляем трекер
ok, bbox = tracker.update(frame)
# Вычисляем fps
fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer)
# рисуем рамку
if ok:
# объект найден
p1 = (int(bbox[0]), int(bbox[1]))
p2 = (int(bbox[0] + bbox[2]), int(bbox[1] + bbox[3]))
cv2.rectangle(frame, p1, p2, (255, 0, 0), 2, 1)
else:
# Объект не найден
cv2.putText(frame, "Cant find any", (100, 80),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)
# HUD трекера( какой трекер используется)
cv2.putText(frame, tracker_type + " Tracker", (100, 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2)
# HUD fps
cv2.putText(frame, "FPS : " + str(int(fps)), (100, 50),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2)
# Вывод
cv2.imshow("Tracking", frame)
# Выход по клавише esc
k = cv2.waitKey(1) & 0xff
if k == 27: break
def init(self, image_path, bbox):
img = cv2.imread(image_path)
self.tracker = cv2.TrackerKCF_create()
self.tracker.init(img, tuple(bbox.to_json()))
self.tracker.update(img)
ap.add_argument("--tracker",
required=False,
help="Path to where the video to track is stored")
args = vars(ap.parse_args())
video = args["image"]
tracker_type = args["tracker"]
cap = cv2.VideoCapture(video)
ret, frame = cap.read()
initBB = cv2.selectROI("Object to track", frame)
cv2.destroyWindow("Object to track")
OPENCV_OBJECT_TRACKERS = {
"csrt": cv2.TrackerCSRT_create(),
"kcf": cv2.TrackerKCF_create(),
"boosting": cv2.TrackerBoosting_create(),
"mil": cv2.TrackerMIL_create(),
"tld": cv2.TrackerTLD_create(),
"medianflow": cv2.TrackerMedianFlow_create(),
"mosse": cv2.TrackerMOSSE_create()
}
if tracker_type:
tracker = OPENCV_OBJECT_TRACKERS[tracker_type]
else:
tracker = cv2.TrackerKCF_create()
tracker.init(frame, initBB)
while cap.isOpened():
ret, frame = cap.read()
def __init__(self,
x_center,
half_width,
y_center,
half_height,
y_bar_start,
half_bar_width,
half_bar_height,
tracker_type='BOOSTING'):
"""
Parameters
----------
x_center : int
Horizontal position of the center of the main area.
half_width : int
Half the width of the main area.
y_center : int
Vertical position of the center of the main area.
half_height : int
Half the height of the main area.
y_bar_start : int
Vertical position of the inner area. This value is relative to the
main area.
half_bar_width : int
Half the width of the inner area.
half_bar_height : int
Half the height of the inner area.
tracker_type : str
Name of one of the built-in trackers in OpenCV.
"""
SectionProcessor.__init__(self)
# Top-left corner of the main area.
self.x_start = int(x_center - half_width)
self.y_start = int(y_center - half_height)
# Bottom-right corner of the main area.
self.x_end = int(x_center + half_width)
self.y_end = int(y_center + half_height)
# Top-left corner of the inner area. This position is relative to the
# main area.
self.y_bar_start = y_bar_start
self.x_bar_start = int(half_width - half_bar_width)
# Inner area
self.bbox = (self.x_bar_start, self.y_bar_start, half_bar_width * 2,
half_bar_height * 2)
# Vertical positions of the top-left corner of the inner area
self.y_pos_history = None
# Indices of the frames where the inner area is in the bottom position.
self.peaks = None
self.events = None
# Tracker
(major_ver, minor_ver, subminor_ver) = (cv.__version__).split('.')
if int(minor_ver) < 3:
self.tracker = cv.Tracker_create(tracker_type)
else:
if tracker_type == 'BOOSTING':
self.tracker = cv.TrackerBoosting_create()
if tracker_type == 'MIL':
self.tracker = cv.TrackerMIL_create()
if tracker_type == 'KCF':
self.tracker = cv.TrackerKCF_create()
if tracker_type == 'TLD':
self.tracker = cv.TrackerTLD_create()
if tracker_type == 'MEDIANFLOW':
self.tracker = cv.TrackerMedianFlow_create()
if tracker_type == 'GOTURN':
self.tracker = cv.TrackerGOTURN_create()
if tracker_type == 'MOSSE':
self.tracker = cv.TrackerMOSSE_create()
if tracker_type == "CSRT":
self.tracker = cv.TrackerCSRT_create()
def trackAllJetsonCam(self,
inputVideo,
save_tracking=False,
show_tracking=True,
Frame_int_Goturn=0,
TimeToRun=30,
camera=False):
"""Track the objects in the video
TimeToRun IN SECONDS !
"""
objRegressor = self.regressor
objTracker = self.tracker
if camera:
vid = cv2.VideoCapture(
"nvcamerasrc ! video/x-raw(memory:NVMM), width=(int)1280, height=(int)720,format=(string)I420, "
"framerate=(fraction)20/1 ! nvvidconv flip-method=0 ! video/x-raw, format=(string)BGRx ! videoconvert ! "
"video/x-raw, format=(string)BGR ! appsink")
num_frames = TimeToRun * 20
else:
vid = cv2.VideoCapture(inputVideo)
num_frames = int(vid.get(cv2.CAP_PROP_FRAME_COUNT))
ok, frame_0 = vid.read()
if not ok:
print('Couldnt read first frame. Exit.')
exit()
FPS = vid.get(cv2.CAP_PROP_FPS)
if save_tracking:
movie_number = 903
tracker_type = 'GoturnGPU'
fourcc = cv2.VideoWriter_fourcc(*'DIVX')
width = int(vid.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(vid.get(cv2.CAP_PROP_FRAME_HEIGHT))
output_video_name = 'output_videos/movie_' + str(
movie_number) + time.strftime(
"_%d%m%Y_%H%M%S") + '_' + tracker_type + '.avi'
outVideo = cv2.VideoWriter(output_video_name, fourcc, int(60),
(width, height))
box_0 = cv2.selectROI(frame_0)
cv2.destroyAllWindows()
bbox_0 = BoundingBox(x1=box_0[0],
y1=box_0[1],
x2=box_0[0] + box_0[2],
y2=box_0[1] + box_0[3])
# Init Goturn
objTracker.init(frame_0, bbox_0, objRegressor)
# Init KCF
multiTracker = cv2.MultiTracker_create()
multiTracker.add(cv2.TrackerKCF_create(), frame_0,
(bbox_0.x1, bbox_0.y1, (bbox_0.x2 - bbox_0.x1),
(bbox_0.y2 - bbox_0.y1)))
num_KCF = 0
bbox = bbox_0
timerI = cv2.getTickCount()
for i in range(2, num_frames):
print('Tracking on frame:' + str(i) + '/' + str(num_frames))
ok, frame = vid.read()
if not ok:
break
### Start timer
timer = cv2.getTickCount()
### Update Tracker
if num_KCF < Frame_int_Goturn:
ok_KCF, box = multiTracker.update(frame)
if ok_KCF:
bbox = BoundingBox(x1=box[0][0],
y1=box[0][1],
x2=box[0][0] + box[0][2],
y2=box[0][1] + box[0][3])
num_KCF += 1
color = (50, 255, 50)
if num_KCF == Frame_int_Goturn or not ok_KCF:
objTracker.update__prev_Jo(frame, bbox)
if not num_KCF < Frame_int_Goturn or not ok_KCF:
num_KCF = 0
bbox = objTracker.track(frame, objRegressor)
multiTracker = cv2.MultiTracker_create()
multiTracker.add(
cv2.TrackerKCF_create(), frame,
(int(bbox.x1), int(bbox.y1), int(bbox.x2 - bbox.x1),
int(bbox.y2 - bbox.y1)))
color = (255, 0, 0)
### Calculate Frames per second (FPS)
fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer)
### Draw bounding box
ImageDraw = frame.copy()
ImageDraw = cv2.rectangle(ImageDraw, (int(bbox.x1), int(bbox.y1)),
(int(bbox.x2), int(bbox.y2)), color, 2)
# Display FPS on frame
cv2.putText(ImageDraw, "FPS : " + str(int(fps)), (100, 50),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (255, 0, 0), 2)
if show_tracking:
cv2.imshow('Results', ImageDraw)
cv2.waitKey(1)
if save_tracking:
outVideo.write(ImageDraw)
timerF = cv2.getTickCount()
Time_proc = (timerF - timerI) / cv2.getTickFrequency()
print('Total processing time =' + str(int(Time_proc)) + 'sec')
print('Average FPS =' + str(i / Time_proc))
cv2.imshow('Results', ImageDraw)
cv2.waitKey(5000)
vid.release()
if save_tracking:
outVideo.release()
def trackAllRealTimeSimu(self,
input_folder,
inputVideo,
save_tracking=False,
show_tracking=True,
Frame_int_Goturn=0):
"""Track the objects in the video
Simulates real time by taking the frame according to computation time of each itteration.
"""
objRegressor = self.regressor
objTracker = self.tracker
vid = cv2.VideoCapture(input_folder + inputVideo)
FPS = vid.get(cv2.CAP_PROP_FPS)
ok, frame_0 = vid.read()
if not ok:
print('Couldnt read first frame. Exit.')
exit()
if save_tracking:
movie_number = 0
tracker_type = 'GoturnGPU'
fourcc = cv2.VideoWriter_fourcc(*'DIVX')
width = int(vid.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(vid.get(cv2.CAP_PROP_FRAME_HEIGHT))
output_video_name = 'output_videos/movie_' + inputVideo + str(
movie_number) + time.strftime(
"_%d%m%Y_%H%M%S") + '_' + tracker_type + '.avi'
outVideo = cv2.VideoWriter(output_video_name, fourcc, int(FPS),
(width, height))
box_0 = cv2.selectROI(frame_0)
cv2.destroyAllWindows()
num_frames = int(vid.get(cv2.CAP_PROP_FRAME_COUNT))
bbox = BoundingBox(x1=box_0[0],
y1=box_0[1],
x2=box_0[0] + box_0[2],
y2=box_0[1] + box_0[3])
# Init Goturn
objTracker.init(frame_0, bbox, objRegressor)
# Init KCF
multiTracker = cv2.MultiTracker_create()
multiTracker.add(cv2.TrackerKCF_create(), frame_0,
(bbox.x1, bbox.y1, (bbox.x2 - bbox.x1),
(bbox.y2 - bbox.y1)))
num_KCF = 0
TimeSimulation = 0
i = 0
color = (50, 255, 50)
while True:
i += 1
# Simulate real time by taking the next frame
TimeSimuFrame = TimeSimulation * FPS
NumframeSimu = int(TimeSimuFrame)
print('Tracking on frame:' + str(NumframeSimu) + '/' +
str(num_frames))
vid.set(1, NumframeSimu)
ok, frame = vid.read()
if not ok:
break
### Start timer
timeA = cv2.getTickCount()
### Update Tracker
if num_KCF < Frame_int_Goturn:
ok_KCF, box = multiTracker.update(frame)
if ok_KCF:
bbox = BoundingBox(x1=box[0][0],
y1=box[0][1],
x2=box[0][0] + box[0][2],
y2=box[0][1] + box[0][3])
num_KCF += 1
if num_KCF == Frame_int_Goturn or not ok_KCF:
objTracker.update__prev_Jo(frame, bbox)
color = (255, 0, 0)
if not num_KCF < Frame_int_Goturn or not ok_KCF:
num_KCF = 0
bbox = objTracker.track(frame, objRegressor)
multiTracker = cv2.MultiTracker_create()
multiTracker.add(
cv2.TrackerKCF_create(), frame,
(int(bbox.x1), int(bbox.y1), int(bbox.x2 - bbox.x1),
int(bbox.y2 - bbox.y1)))
bbox = objTracker.track(frame, objRegressor)
### Calculate Frames per second (FPS)
timeB = cv2.getTickCount()
fps = cv2.getTickFrequency() / (timeB - timeA)
TimeSimulation += 1 / fps
### Draw bounding box
ImageDraw = frame.copy()
ImageDraw = cv2.rectangle(ImageDraw, (int(bbox.x1), int(bbox.y1)),
(int(bbox.x2), int(bbox.y2)), color, 2)
color = (50, 255, 50)
# Display FPS on frame
cv2.putText(ImageDraw, "FPS : " + str(int(fps)), (100, 50),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (255, 0, 0), 2)
if show_tracking:
cv2.imshow('Results', ImageDraw)
cv2.waitKey(1)
if save_tracking:
outVideo.write(ImageDraw)
timerF = cv2.getTickCount()
print('Total processing time =' + str(TimeSimulation))
print('Average FPS =' + str(i / TimeSimulation))
if save_tracking:
outVideo.release()
def trackAllGoturnKcf(self,
inputVideo,
save_tracking=False,
show_tracking=True,
Frame_int_Goturn=0):
"""Track the objects in the video
Combining Opencv's KCF tracking with the goturn tracking.
"""
objRegressor = self.regressor
objTracker = self.tracker
vid = cv2.VideoCapture(inputVideo)
ok, frame_0 = vid.read()
if not ok:
print('Couldnt read first frame. Exit.')
exit()
if save_tracking:
movie_number = 0
tracker_type = 'GoturnGPU'
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
width = int(vid.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(vid.get(cv2.CAP_PROP_FRAME_HEIGHT))
FPS = vid.get(cv2.CAP_PROP_FPS)
output_video_name = 'output_videos/movie_' + str(
movie_number) + time.strftime(
"_%d%m%Y_%H%M%S") + '_' + tracker_type + '.mp4'
outVideo = cv2.VideoWriter(output_video_name, fourcc, int(FPS),
(width, height))
box_0 = cv2.selectROI(frame_0)
cv2.destroyAllWindows()
num_frames = int(vid.get(cv2.CAP_PROP_FRAME_COUNT))
bbox_0 = BoundingBox(x1=box_0[0],
y1=box_0[1],
x2=box_0[0] + box_0[2],
y2=box_0[1] + box_0[3])
# Init Goturn
objTracker.init(frame_0, bbox_0, objRegressor)
# Init KCF
multiTracker = cv2.MultiTracker_create()
multiTracker.add(cv2.TrackerKCF_create(), frame_0,
(bbox_0.x1, bbox_0.y1, (bbox_0.x2 - bbox_0.x1),
(bbox_0.y2 - bbox_0.y1)))
num_KCF = 0
bbox = bbox_0
timerI = cv2.getTickCount()
for i in range(2, num_frames):
print('Tracking on frame:' + str(i) + '/' + str(num_frames))
ok, frame = vid.read()
if not ok:
break
### Start timer
timer = cv2.getTickCount()
### Update Tracker
if num_KCF < Frame_int_Goturn:
ok_KCF, box = multiTracker.update(frame)
if ok_KCF:
bbox = BoundingBox(x1=box[0][0],
y1=box[0][1],
x2=box[0][0] + box[0][2],
y2=box[0][1] + box[0][3])
num_KCF += 1
color = (50, 255, 50)
if num_KCF == Frame_int_Goturn or not ok_KCF:
objTracker.update__prev_Jo(frame, bbox)
if not num_KCF < Frame_int_Goturn or not ok_KCF:
num_KCF = 0
bbox = objTracker.track(frame, objRegressor)
multiTracker = cv2.MultiTracker_create()
multiTracker.add(
cv2.TrackerKCF_create(), frame,
(int(bbox.x1), int(bbox.y1), int(bbox.x2 - bbox.x1),
int(bbox.y2 - bbox.y1)))
color = (255, 0, 0)
### Calculate Frames per second (FPS)
fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer)
### Draw bounding box
ImageDraw = frame.copy()
ImageDraw = cv2.rectangle(ImageDraw, (int(bbox.x1), int(bbox.y1)),
(int(bbox.x2), int(bbox.y2)), color, 2)
# Display FPS on frame
cv2.putText(ImageDraw, "FPS : " + str(int(fps)), (100, 50),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (255, 0, 0), 2)
if show_tracking:
cv2.imshow('Results', ImageDraw)
cv2.waitKey(1)
if save_tracking:
outVideo.write(ImageDraw)
timerF = cv2.getTickCount()
Time_proc = (timerF - timerI) / cv2.getTickFrequency()
print('Total processing time =' + str(int(Time_proc)) + 'sec')
print('Average FPS =' + str(i / Time_proc))
cv2.imshow('Results', ImageDraw)
cv2.waitKey(5000)
vid.release()
if save_tracking:
outVideo.release()
def __init__(self, frame, face):
(x, y, w, h) = face
self.face = (x, y, w, h)
# Arbitrarily picked KCF tracking
self.tracker = cv2.TrackerKCF_create()
self.tracker.init(frame, self.face)
def main():
"""Control function that reads webcam, and tracks a marked object."""
global x0, y0, x1, y1, drawing, mode, frame, bbox, tracker, tracker_initialized
global MODE_MARK
#
# initialization
#
(major_ver, minor_ver, subminor_ver) = (cv2.__version__).split('.')
MODE_TRACK = 0 # track an object
MODE_SHOW = 1 # only show tracking markers on video
MODE_MARK = 2 # mark region color to track
MODE_TRACK_HOLD = 3 # temporarily suspend tracking until object is recaptured
tracker_initialized = False
bbox = (0, 0, 0, 0)
last_good_bbox = bbox
mode = MODE_SHOW
mode_text = 'Show'
fps_text = '?? Fps'
batt_level = 0.0
batt_level_text = '?? [v]'
cvs_title_printed = False
drawing = False # true if mouse is pressed
x0, y0 = -1, -1
x1, y1 = -1, -1
#
# PID constants for pan and tilt PID controller.
# These constants are tuned for the worm-gear mechanical setup
#
pan_P = 0.8
pan_I = 0.05
pan_D = 0.1
tilt_P = 0.6
tilt_I = 0.05
tilt_D = 0.1
print ' m - mark color region to track\n t - track\n s - display tracking marker only\n ESC - quit'
#
# connect to NXT device
# define motor references
#
devNXT = nxt.locator.find_one_brick()
tilt_motor = nxt.motor.Motor(devNXT, nxt.motor.PORT_A)
pan_motor = nxt.motor.Motor(devNXT, nxt.motor.PORT_B)
#
# Initialize start time and frame count.
# Initialize a reference start time for the CSV output trace
#
frame_count = 0
start = time.time()
ref_time = time.time()
#
# link event callback function
#
cv2.namedWindow('image', cv2.WINDOW_GUI_NORMAL+cv2.WINDOW_AUTOSIZE)
cv2.setMouseCallback('image', mark_rect)
#
# setup font for overlay text
#
font = cv2.FONT_HERSHEY_SIMPLEX
#
# Set up tracker.
# Tracker algorithm is hard coded here to default tracker KCF.
#
tracker_types = ['BOOSTING', 'MIL','KCF', 'TLD', 'MEDIANFLOW', 'GOTURN']
tracker_type = tracker_types[2]
if int(minor_ver) < 3:
tracker = cv2.Tracker_create(tracker_type)
else:
if tracker_type == 'BOOSTING':
tracker = cv2.TrackerBoosting_create()
if tracker_type == 'MIL':
tracker = cv2.TrackerMIL_create()
if tracker_type == 'KCF':
tracker = cv2.TrackerKCF_create()
if tracker_type == 'TLD':
tracker = cv2.TrackerTLD_create()
if tracker_type == 'MEDIANFLOW':
tracker = cv2.TrackerMedianFlow_create()
if tracker_type == 'GOTURN':
tracker = cv2.TrackerGOTURN_create()
#
# Open the capture device and print some
# useful properties.
# This tracker program will leave the default webcam frame size
# that is 640x480 for the webcam I am using.
#
cap = cv2.VideoCapture(0)
if cap.isOpened():
#cap.set(cv.CV_CAP_PROP_FRAME_WIDTH, 320)
#cap.set(cv.CV_CAP_PROP_FRAME_HEIGHT, 240)
frameWidth = cap.get(cv.CV_CAP_PROP_FRAME_WIDTH)
frameHeight = cap.get(cv.CV_CAP_PROP_FRAME_HEIGHT)
print 'frame: width {}, height {}'.format(frameWidth, frameHeight)
frameCenterX = int(frameWidth/2)
frameCenterY = int(frameHeight/2)
else:
sys.exit()
#
# frame capture and processing loop
#
while(True):
#
# capture a frame
# cover to appropriate color space to improve detection
# in different lighting conditions
#
cap_ok, frame = cap.read()
if not cap_ok:
break
#
# Operations on the captured image done here.
# If marking a section on the frame for tracking
# then only display the ROI selection
#
if mode == MODE_MARK:
cv2.rectangle(frame, (x0, y0), (x1, y1), (0, 255, 0), 1)
#
# If tracking or only displaying object tracking information
# then draw the tracking markers on the frame before it is displayed.
# Only do this if the tracker was initialized
#
elif tracker_initialized:
#
# Update the tracker with the newly acquired frame.
#
track_ok, bbox = tracker.update(frame)
#
# If the tracker update was successful, object still being tracked, then
# update the prev bounding box position and proceed to:
# - display the tracker bounding box
# - an arrow line from frame center to the object center
# - calculate pan and tilt error from frame center
#
if track_ok:
last_good_bbox = bbox
p1 = (int(bbox[0]), int(bbox[1]))
p2 = (int(bbox[0] + bbox[2]), int(bbox[1] + bbox[3]))
cv2.rectangle(frame, p1, p2, (255, 0, 0), 2)
object_x = int(bbox[0] + bbox[2]/2)
object_y = int(bbox[1] + bbox[3]/2)
err_pan_i = frameCenterX - object_x
err_tilt_i = frameCenterY - object_y
cv2.arrowedLine(frame, (frameCenterX, frameCenterY), (object_x, object_y), (255, 0, 0), 2)
if mode == MODE_TRACK_HOLD:
mode = MODE_TRACK
#
# If tracking is lost for some reason then use the last location
# of the bounding box to mark that last location with a red box
#
else:
p1 = (int(last_good_bbox[0]), int(last_good_bbox[1]))
p2 = (int(last_good_bbox[0] + last_good_bbox[2]), int(last_good_bbox[1] + last_good_bbox[3]))
cv2.rectangle(frame, p1, p2, (0, 0, 255), 1)
if mode == MODE_TRACK:
mode = MODE_TRACK_HOLD
#
# Only when in tracking mode activate the motors,
# and use PID calculations to control the pan-tilt device
#
if mode == MODE_TRACK and tracker_initialized:
#
# First apply an exponential filter to the tracker position error.
# info: https://en.wikipedia.org/wiki/Exponential_smoothing
# Then do PID tracking for NXT motors' pan-tilt control.
#
err_pan = exp_filter_pan(err_pan_i)
err_tilt = exp_filter_tilt(err_tilt_i)
control_pan = pid_pan(err_pan, pan_P, pan_I, pan_D)
control_tilt = -1.0 * pid_tilt(err_tilt, tilt_P, tilt_I, tilt_D)
#
# Print out some data in a CSV compatible format for graphing
#
if not cvs_title_printed:
print 'rel_time,err_tilt_i,err_tilt,control_tilt,err_pan_i,err_pan,control_pan'
cvs_title_printed = True
now = time.time() - ref_time
print '{:.2f},{:.2f},{:.2f},{:.2f},{:.2f},{:.2f},{:.2f}'.format(now,err_tilt_i, err_tilt, control_tilt, err_pan_i, err_pan, control_pan)
#
# activate NXT motors
#
control_pan = power_limit(control_pan, 90.0)
control_tilt = power_limit(control_tilt, 90.0)
# uncomment one of the following lines
# in order to isolate pan or tilt for PID tuning/testing
#control_pan = 0.0
#control_tilt = 0.0
pan_motor.run(int(control_pan), True)
tilt_motor.run(int(control_tilt), True)
#
# This section will turn motors off
# when not in tracking mode. Note above code lines
# force MODE_TRACK_HOLD if no objects exist or if more than one
# object is detected. This state will shut motors off.
#
else:
pan_motor.idle()
tilt_motor.idle()
#
# Calculate and display FPS.
# Use the 10sec interval to also poll the NXT for battery level.
#
frame_count = frame_count + 1
end = time.time()
measure_interval = end - start
if measure_interval > 10:
fps = frame_count / measure_interval
fps_text = '{:.2f} Fps'.format(fps)
frame_count = 0
start = time.time()
batt_level = devNXT.get_battery_level() / 1000.0
batt_level_text = '{:.2f} [v]'.format(batt_level)
#
# Add text and other markers to the image
#
if mode == MODE_TRACK_HOLD:
cv2.putText(frame, mode_text, (1, 20), font, 0.4, (0, 0, 255), 1, cv2.LINE_AA)
else:
cv2.putText(frame, mode_text, (1, 20), font, 0.4, (255, 0, 0), 1, cv2.LINE_AA)
cv2.putText(frame, tracker_type, (1, 40), font, 0.4, (255, 0, 0), 1, cv2.LINE_AA)
cv2.putText(frame, fps_text, (1, 60), font, 0.4, (255, 0, 0), 1, cv2.LINE_AA)
if batt_level > 6.50:
cv2.putText(frame, batt_level_text, (1, 80), font, 0.4, (255, 0, 0), 1, cv2.LINE_AA)
else:
cv2.putText(frame, batt_level_text, (1, 80), font, 0.4, (0, 0, 255), 1, cv2.LINE_AA)
#
# Display the resulting frame
#
cv2.imshow('image', frame)
#
# key input mode/command
#
key = cv2.waitKey(1) & 0xFF
if key == 27:
break
elif key == ord('m'):
x0,y0 = -1,-1
x1,y1 = -1,-1
mode_text = 'Mark'
mode = MODE_MARK
elif key == ord('t'):
mode_text = 'Track'
if tracker_initialized:
mode = MODE_TRACK
else:
mode = MODE_TRACK_HOLD
elif key == ord('s'):
mode_text = 'Show'
mode = MODE_SHOW
else:
pass
#
# When done, stop motors and release the capture.
#
pan_motor.idle()
tilt_motor.idle()
cap.release()
cv2.destroyAllWindows()
def object_track():
# Set up tracker.
# Instead of MIL, you can also use
tracker_types = [
'BOOSTING', 'MIL', 'KCF', 'TLD', 'MEDIANFLOW', 'GOTURN', 'MOSSE',
'CSRT'
]
tracker_type = tracker_types[0]
if int(minor_ver) < 3:
tracker = cv2.Tracker_create(tracker_type)
else:
if tracker_type == 'BOOSTING':
tracker = cv2.TrackerBoosting_create()
if tracker_type == 'MIL':
tracker = cv2.TrackerMIL_create()
if tracker_type == 'KCF':
tracker = cv2.TrackerKCF_create()
if tracker_type == 'TLD':
tracker = cv2.TrackerTLD_create()
if tracker_type == 'MEDIANFLOW':
tracker = cv2.TrackerMedianFlow_create()
if tracker_type == 'GOTURN':
tracker = cv2.TrackerGOTURN_create()
if tracker_type == 'MOSSE':
tracker = cv2.TrackerMOSSE_create()
if tracker_type == "CSRT":
tracker = cv2.TrackerCSRT_create()
# Read video
video = cv2.VideoCapture(0)
# Exit if video not opened.
if not video.isOpened():
print "Could not open video"
sys.exit()
# Read first frame.
ok, frame = video.read()
if not ok:
print 'Cannot read video file'
sys.exit()
# Define an initial bounding box
bbox = (287, 23, 86, 320)
# Uncomment the line below to select a different bounding box
bbox = cv2.selectROI(frame, False)
cv2.destroyAllWindows()
# Initialize tracker with first frame and bounding box
while (True):
ret, frame = video.read()
if ret:
k = cv2.waitKey(30)
p1 = (int(bbox[0]), int(bbox[1]))
p2 = (int(bbox[0] + bbox[2]), int(bbox[1] + bbox[3]))
cv2.rectangle(frame, p1, p2, (255, 0, 0), 2, 1)
frame = cv2.flip(frame, 1)
cv2.imshow("Initialize Tracking", frame)
if k == 32:
break
cv2.destroyAllWindows()
ok = tracker.init(frame, bbox)
while True:
# Read a new frame
ok, frame = video.read()
if not ok:
break
# Start timer
timer = cv2.getTickCount()
# Update tracker
ok, bbox = tracker.update(frame)
# Calculate Frames per second (FPS)
fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer)
# Draw bounding box
if ok:
# Tracking success
p1 = (int(bbox[0]), int(bbox[1]))
p2 = (int(bbox[0] + bbox[2]), int(bbox[1] + bbox[3]))
cv2.rectangle(frame, p1, p2, (255, 0, 0), 2, 1)
else:
# Tracking failure
cv2.putText(frame, "Tracking failure detected", (100, 80),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)
# Display tracker type on frame
cv2.putText(frame, tracker_type + " Tracker", (100, 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2)
# Display FPS on frame
cv2.putText(frame, "FPS : " + str(int(fps)), (100, 50),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2)
# Display result
frame = cv2.flip(frame, 1)
cv2.imshow("Tracking", frame)
# Exit if ESC pressed
k = cv2.waitKey(30) & 0xff
if k == 27: break
'BOOSTING', 'MIL', 'KCF', 'TLD', 'MEDIANFLOW', 'GOTURN', 'MOSSE', 'CSRT'
]
tracker_type = tracker_types[7]
if int(major_ver) < 3:
finger_tracker = cv2.Tracker_create(tracker_type)
phone_tracker = cv2.Tracker_create(tracker_type)
else:
if tracker_type == 'BOOSTING':
finger_tracker = cv2.TrackerBoosting_create()
phone_tracker = cv2.TrackerBoosting_create()
if tracker_type == 'MIL':
finger_tracker = cv2.TrackerMIL_create()
phone_tracker = cv2.TrackerMIL_create()
if tracker_type == 'KCF':
finger_tracker = cv2.TrackerKCF_create()
phone_tracker = cv2.TrackerKCF_create()
if tracker_type == 'TLD':
finger_tracker = cv2.TrackerTLD_create()
phone_tracker = cv2.TrackerTLD_create()
if tracker_type == 'MEDIANFLOW':
finger_tracker = cv2.TrackerMedianFlow_create()
phone_tracker = cv2.TrackerMedianFlow_create()
if tracker_type == 'GOTURN':
finger_tracker = cv2.TrackerGOTURN_create()
phone_tracker = cv2.TrackerGOTURN_create()
if tracker_type == 'MOSSE':
finger_tracker = cv2.TrackerMOSSE_create()
phone_tracker = cv2.TrackerMOSSE_create()
if tracker_type == "CSRT":
finger_tracker = cv2.TrackerCSRT_create()
def __init__(self, flag):
self.batch_size = 32 #How many experiences to use for each training step.
self.update_freq = 4 #How often to perform a training step.
self.y = .99 #Discount factor on the target Q-values
self.startE = 1 #Starting chance of random action
self.endE = 0.1 #Final chance of random action
self.annealing_steps = 10000. #How many steps of training to reduce startE to endE.
self.num_episodes = 10000 #How many episodes of game environment to train network with.
self.pre_train_steps = 10000 #How many steps of random actions before training begins.
self.max_epLength = 500 #The max allowed length of our episode.
self.load_model = False #Whether to load a saved model.
self.path = "./dqn" #The path to save our model to.
self.h_size = 512 #The size of the final convolutional layer before splitting it into Advantage and Value streams.
self.tau = 0.001 #Rate to update target network toward primary network
self.action_num = 5
tf.reset_default_graph()
self.mainQN = DQN(self.h_size, self.action_num)
self.targetQN = DQN(self.h_size, self.action_num)
self.init = tf.global_variables_initializer()
self.saver = tf.train.Saver()
self.trainables = tf.trainable_variables()
self.targetOps = self.updateTargetGraph(self.trainables, self.tau)
self.myBuffer = experience_buffer()
# Set the rate of random action decrease.
self.e = self.startE
self.stepDrop = (self.startE - self.endE) / self.annealing_steps
# create lists to contain total rewards and steps per episode
self.jList = []
self.rList = []
self.total_steps = 0
self.game = sim(200, True)
self.is_Train = flag
# for Tracking
self.cap = None
self.col = -1
self.width = -1
self.row = -1
self.height = -1
self.frame = None
self.frame2 = None
self.inputmode = False
self.rectangle = False
self.trackWindow = None
self.roi_hist = None
self.roi = None
self.caffe_model_path = './MobileNetSSD_deploy.caffemodel'
self.prorotxt_path = './MobileNetSSD_deploy.prototxt.txt'
self.net = None
self.obstacle_points = []
self.target_point = None
self.obstacle_box_color = (0, 0, 255)
self.tracker_types = [
'BOOSTING', 'MIL', 'KCF', 'TLD', 'MEDIANFLOW', 'GOTURN'
]
self.tracker_type = self.tracker_types[2]
if self.tracker_type == 'BOOSTING':
self.tracker = cv2.TrackerBoosting_create()
if self.tracker_type == 'MIL':
self.tracker = cv2.TrackerMIL_create()
if self.tracker_type == 'KCF':
self.tracker = cv2.TrackerKCF_create()
if self.tracker_type == 'TLD':
self.tracker = cv2.TrackerTLD_create()
else:
self.tracker = cv2.TrackerMedianFlow_create()
def cv_tracking():
tracker_types = ['BOOSTING', 'MIL', 'KCF', 'TLD', 'MEDIANFLOW', 'GOTURN']
tracker_type = tracker_types[0]
if tracker_type == 'BOOSTING':
tracker = cv2.TrackerBoosting_create()
if tracker_type == 'MIL':
tracker = cv2.TrackerMIL_create()
if tracker_type == 'KCF':
tracker = cv2.TrackerKCF_create()
if tracker_type == 'TLD':
tracker = cv2.TrackerTLD_create()
if tracker_type == 'MEDIANFLOW':
tracker = cv2.TrackerMedianFlow_create()
if tracker_type == 'GOTURN':
tracker = cv2.TrackerGOTURN_create()
# Read video
video = cv2.VideoCapture("videos/car-without_obstacle_300.avi")
ok, frame = video.read()
# check for video
if not video.isOpened():
print("video not opened")
sys.exit()
# Read first frame.
count = 0
while (count < 290):
ok, frame = video.read()
if not ok:
print('Cannot read video file')
sys.exit()
count += 1
# Define an initial bounding box
# bbox = (287, 23, 86, 320)
# Uncomment the line below to select a different bounding box
bbox = cv2.selectROI(frame, False)
# Initialize tracker with first frame and bounding box
ok = tracker.init(frame, bbox)
while True:
# Read a new frame
ok, frame = video.read()
if not ok:
break
# Start timer
timer = cv2.getTickCount()
# Update tracker
ok, bbox = tracker.update(frame)
print(bbox)
# Calculate Frames per second (FPS)
fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer)
# Draw bounding box
if ok:
# Tracking success
p1 = (int(bbox[0]), int(bbox[1]))
p2 = (int(bbox[0] + bbox[2]), int(bbox[1] + bbox[3]))
cv2.rectangle(frame, p1, p2, (255, 0, 0), 2, 1)
else:
# Tracking failure
cv2.putText(frame, "Tracking failure detected", (100, 80),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)
# Display tracker type on frame
cv2.putText(frame, tracker_type + " Tracker", (100, 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2)
# Display FPS on frame
cv2.putText(frame, "FPS : " + str(int(fps)), (100, 50),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2)
# Display result
cv2.imshow("Tracking", frame)
# Exit if ESC pressed
k = cv2.waitKey(1) & 0xff
if k == 27: break
def trackFaces(frame, faces, exitFlag, facesNumber, boxes, receiveQueue,
sendQueue):
boxes = boxes
while (facesNumber != 0 and exitFlag == 0):
# Create tracker
tracker = cv2.MultiTracker_create()
for face in faces:
ok = tracker.add(
cv2.TrackerKCF_create(), frame,
(int(face[0]), int(face[1]), int(face[2]), int(face[3])))
#time = cv2.getTickCount()
while (exitFlag == 0):
#print("Tracking...")
# Start timer
timer = cv2.getTickCount()
# Update tracker
ret, boxes = tracker.update(frame)
# Calculate Frames per second (FPS)
fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer)
if ret: # Tracking success
for box in boxes:
topLeft = (int(box[0]), int(box[1]))
bottomDown = (int(box[0] + box[2]), int(box[1] + box[3]))
cv2.rectangle(frame, topLeft, bottomDown, (255, 0, 0), 2,
1) # Draw box
else:
break
# Display tracker type on frame
cv2.putText(frame, "KCF Tracker", (100, 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2)
# Display FPS on frame
cv2.putText(frame, "FPS : " + str(int(fps)), (100, 50),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2)
# Display result
cv2.imshow("Tracking", frame)
# Read frame
ok, frame = video.read()
frame = cv2.flip(frame, 1) #flip image to adjust it
if (sendQueue.empty() == True
): #if detector did not consume the frame continue
sendQueue.put([frame, faces, exitFlag, facesNumber])
# Exit if ESC pressed
k = cv2.waitKey(1) & 0xff
if k == 27:
exitFlag = 1
sendQueue.put([frame, faces, exitFlag, facesNumber])
break
if (receiveQueue.empty() == 0): # new face
break
print("Exited Tracking...")
# Read frame
ok, frame = video.read()
frame = cv2.flip(frame, 1) #flip image to adjust it
if (sendQueue.empty() == True):
sendQueue.put([frame, faces, exitFlag, facesNumber])
# Exit if ESC pressed
k = cv2.waitKey(1) & 0xff
if k == 27:
exitFlag = 1
sendQueue.put([frame, faces, exitFlag, facesNumber])
break
for box in boxes:
topLeft = (int(box[0]), int(box[1]))
bottomDown = (int(box[0] + box[2]), int(box[1] + box[3]))
cv2.rectangle(frame, topLeft, bottomDown, (255, 0, 0), 2,
1) # Draw box
# Display result
cv2.imshow("Tracking", frame)
if (receiveQueue.empty() == 0): # new face
break
# Read frame
ok, frame = video.read()
frame = cv2.flip(frame, 1) #flip image to adjust it
if (sendQueue.empty() == True):
sendQueue.put([frame, faces, exitFlag, facesNumber])
# Exit if ESC pressed
k = cv2.waitKey(1) & 0xff
if k == 27:
exitFlag = 1
sendQueue.put([frame, faces, exitFlag, facesNumber])
for box in boxes:
topLeft = (int(box[0]), int(box[1]))
bottomDown = (int(box[0] + box[2]), int(box[1] + box[3]))
cv2.rectangle(frame, topLeft, bottomDown, (255, 0, 0), 2,
1) # Draw box
# Display result
cv2.imshow("Tracking", frame)
return frame, faces, exitFlag, facesNumber, boxes
def background_subtract(video):
frameCounter = 0
trackerArray = []
#kalmanArray = []
duplicate = False
fgbg = cv2.createBackgroundSubtractorMOG2(
detectShadows=False, history=200,
varThreshold=200) #Starting back ground subtractor object
while (1):
#Loading video, one frame at a time
ret, frame = video.read()
#transform the picture
#frame = transform_perspective(frame)
#Blurring and applying erosion and dilation to the element used for blob detection
fgmask = fgbg.apply(frame)
blur_size = 3
cv2.blur(frame, (blur_size, blur_size), frame)
elip_val = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (5, 5))
fgmask = cv2.erode(fgmask, elip_val)
fgmask = cv2.dilate(fgmask, elip_val, iterations=5)
#Finding new blobs each 30 frames
if frameCounter % 200 == 0 or frameCounter == 2:
print("200 frames passed, updating blobs")
centroids = blob_detection('test', fgmask, fgmask)
#Going through all the centroids and comparing them with the tracked blobs, not adding new ones whichs is too close to an existing tracked object
for i in centroids:
one_bbox = i
bbox = (one_bbox[0] - 20, one_bbox[1] - 20, 40, 40)
duplicate = False
for j in range(len(trackerArray)):
ok, t = trackerArray[j].update(frame)
if sqrt((bbox[0] - t[0])**2 +
(bbox[1] - t[1])**2) < 50 and ok:
duplicate = True
if duplicate == False:
#Creating a new tracker and a new kalman filter for the blob
tracker = cv2.TrackerKCF_create()
ok = tracker.init(frame, bbox)
#kFilter = Kalman(bbox[0], bbox[1])
trackerArray.append(tracker)
#kalmanArray.append(kFilter)
'''
# Update tracker & Kalman filter
cnt = 0
for i in trackerArray:
ok, bbox = i.update(frame)
kMan = kalmanArray[cnt]
#Calculating velocity from old positional values and time since last measurements
deltaX = bbox[0] - kMan.getX()
deltaY = bbox[1] - kMan.getY()
xVel = deltaX / 0.04 # 1/25 which is the time in seconds from last frame
yVel = deltaY / 0.04
kMan.update(bbox[0], bbox[1], xVel, yVel)
estimate = kMan.estimate()
if ok:
# Tracking success
p1 = (int(bbox[0]), int(bbox[1]))
p2 = (int(bbox[0] + bbox[2]), int(bbox[1] + bbox[3]))
cv2.rectangle(frame, p1, p2, (255,0,0), 2, 1)
cv2.rectangle(frame, (estimate[0], estimate[1]), (estimate[0]+10, estimate[1]+10), (0,255,0), 2, 1)
distance = sqrt(estimate[2]**2+estimate[3]**2)
#Going from pixels to km/h
distance = (distance/10.1159156)*3.6
cv2.putText(frame, str(float(("%.2f" % distance))) + "km/h", (int(bbox[0]), int(bbox[1])), cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50,170,50), 2)
else:
del trackerArray[cnt]
del kalmanArray[cnt]
cnt = cnt + 1
'''
frameCounter = frameCounter + 1
# Display result
cv2.imshow("Tracking", frame)
cv2.imshow("blobs", fgmask)
# Exit if q pressed
if cv2.waitKey(25) & 0xFF == ord('q'):
break
video.release()
cv2.destroyAllWindows()
import cv2
import sys
import time
tracker = cv2.TrackerKCF_create() # KCF fast and accurate, CSRT high accurate but slower, MOSSE faster but low accuracy
# For video file
video = cv2.VideoCapture("./chaplin.mp4")
# for video stream / webcam
# video = cv2.VideoStream(src=0).start()
# time.sleep(2.0)
ok, frame = video.read()
bbox = cv2.selectROI(frame, False)
print(bbox)
#cv2.destroyWindow(frame)
tracker.init(frame, bbox)
while True:
ok, frame = video.read()
timer = cv2.getTickCount()
ok, bbox = tracker.update(frame)
fps = cv2.getTickFrequency() / (cv2.getTickCount()-timer)
p1 = (int(bbox[0]), int(bbox[1]))
p2 = (int(bbox[0]+bbox[2]), int(bbox[1]+bbox[3]))
cv2.rectangle(frame, p1, p2, (255, 0, 0), 2, 1)
}
number = input(
"1. CSRT 2. KCF\n3. TLD 4. MIL\n5. MOSSE 6. Boosting\n7. MedianFlow\n>> "
)
try:
print("\n追蹤算法 :", method[number])
name = input("影片名稱 : ")
filetype = input("影片格式 : ")
'''
if elif else 單行寫法
True1 if Condition1 else True2 if Condition2 else True3 ... if Condition#n-1 else True#n
'''
tracker = cv2.TrackerCSRT_create() if int(
number) == 1 else cv2.TrackerKCF_create() if int(
number) == 2 else cv2.TrackerTLD_create() if int(
number) == 3 else cv2.TrackerMIL_create(
) if int(number) == 4 else cv2.TrackerMOSSE_create(
) if int(number) == 5 else cv2.TrackerBoosting_create(
) if int(number) == 6 else cv2.TrackerMedianFlow_create()
#print(type(tracker))
#tracker = cv2.TrackerCSRT_create()
#tracker = cv2.TrackerKCF_create()
#tracker = cv2.TrackerTLD_create()
#tracker = cv2.TrackerMIL_create()
#tracker = cv2.TrackerMOSSE_create()
#tracker = cv2.TrackerBoosting_create()
#tracker = cv2.TrackerMedianFlow_create()
else:
return #print("il target si trova di fronte al robot")
else:
return print("il target è vicino")
def tracking_Qr(
frame): #nel momento in cui riconosco un qr code inizio a tracciarlo
barcode = decode(frame)
if (len(barcode) == 1):
for obj in barcode:
box = [obj.rect[0], obj.rect[1], obj.rect[2], obj.rect[3]]
return box #la funzione restituisce le dimensioni del box tracker
tracker = cv2.TrackerKCF_create() #inizializzo un tracker(MedianFlow)
bbox = (20, 30, 50, 60)
cap = cv2.VideoCapture(0) #accedo allo stream della videocamera
if (cap.isOpened() == False): #controllo che il programma usi la videocamera
print("Error opening video stream or file"
) #cap.isOpened() riporta True se è attiva la videocamera
lista = get_size_cam(cap) #creo una lista con altezza e larghezza del frame
print("DIMENSIONE FRAME altezza= ", lista[0], " larghezza= ", lista[1])
while (cap.isOpened()):
ret, frame = cap.read(
) #Ret riporta true se ho un frame in ingresso. Frame rappresenta la matrice dell'immagine
gray_img = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
info = get_barcode_info(
def main():
tracker_types = ['BOOSTING', 'MIL', 'KCF', 'TLD', 'MEDIANFLOW', 'GOTURN', 'MOSSE', 'CSRT']
tracker_type = tracker_types[2]
if int(minor_ver) < 3:
tracker = cv2.Tracker_create(tracker_type)
else:
if tracker_type == 'BOOSTING':
tracker = cv2.TrackerBoosting_create()
if tracker_type == 'MIL':
tracker = cv2.TrackerMIL_create()
if tracker_type == 'KCF':
tracker = cv2.TrackerKCF_create()
if tracker_type == 'TLD':
tracker = cv2.TrackerTLD_create()
if tracker_type == 'MEDIANFLOW':
tracker = cv2.TrackerMedianFlow_create()
if tracker_type == 'GOTURN':
tracker = cv2.TrackerGOTURN_create()
if tracker_type == 'MOSSE':
tracker = cv2.TrackerMOSSE_create()
if tracker_type == "CSRT":
tracker = cv2.TrackerCSRT_create()
x = 200
y = 200
w = 224
h = 224
track_window=(x,y,w,h)
# Reference Distance
L0 = 100
S0 = 50176 #224x224 #take#here.
# Base Distance
LB = 100
# Define an initial bounding box
bbox = (x, y, w, h) #(287, 23, 86, 320)
#CX=int(bbox[0]+0.5*bbox[2]+3) #adding
#CY=int(bbox[1]+0.5*bbox[3]+3) #adding
drone = tellopy.Tello()
try:
drone.connect()
drone.wait_for_connection(60.0)
retry = 3
container = None
while container is None and 0 < retry:
retry -= 1
try:
container = av.open(drone.get_video_stream())
except av.AVError as ave:
print(ave)
print('retry...')
#drone.takeoff()
#sleep(5)
#drone.land()
# skip first 300 frames
frame_skip = 300
while True:
#------------------------------------------for start
for frame in container.decode(video=0):
if 0 < frame_skip:
frame_skip = frame_skip - 1
continue
image = cv2.cvtColor(numpy.array(frame.to_image()), cv2.COLOR_RGB2BGR)
start_time = time.time()
#cv2.imshow('Canny', cv2.Canny(image, 100, 200))
#cv2.waitKey(1)
if frame.time_base < 1.0/60:
time_base = 1.0/60
else:
time_base = frame.time_base
frame_skip = int((time.time() - start_time)/time_base)
# Update tracker
ok, bbox = tracker.update(image)
# Draw bounding box
if ok:
print('Tracking ok')
#(x,y,w,h) = (int(bbox[0]),int(bbox[1]),int(bbox[2]),int(bbox[3]))
CX=int(bbox[0]+0.5*bbox[2]) #Center of X
CY=int(bbox[1]+0.5*bbox[3])
S0=bbox[2]*bbox[3]
print("CX,CY,S0,x,y=",CX,CY,S0,x,y)
# Tracking success
p1 = (x, y)
p2 = (x + w, y + h)
cv2.rectangle(image, p1, p2, (255,0,0), 2, 1)
else:
# Tracking failure
print('Tracking failure')
cv2.putText(image, "Tracking failure detected", (100,80), cv2.FONT_HERSHEY_SIMPLEX, 0.75,(0,0,255),2)
cv2.imshow('Original', image)
key = cv2.waitKey(1)&0xff
if key == ord('q'):
print('Q!')
break
if key == ord('r'):
bbox = cv2.selectROI(image, False)
print(bbox)
(x,y,w,h) = (int(bbox[0]),int(bbox[1]),int(bbox[2]),int(bbox[3]))
# Initialize tracker with first frame and bounding box
ok = tracker.init(image, bbox)
#-------------------------------------------------for end
break
print('stop fly')
except Exception as ex:
exc_type, exc_value, exc_traceback = sys.exc_info()
traceback.print_exception(exc_type, exc_value, exc_traceback)
print(ex)
finally:
drone.quit()
drone.land()
cv2.destroyAllWindows()
true_w, true_h, true_coords = extractor.get_coords(
extractor.text_list[index][0])
# videos_per_person = extractor.video_per_name[index]
print(extractor.text_list[index][0])
print(len(true_coords))
multiTracker = cv2.MultiTracker_create()
multiTracker_body = cv2.MultiTracker_create()
face_boxes, names = recognize_faces(frame, frameWidth, frameHeight,
true_w[frame_number],
true_h[frame_number])
body_boxes, openpose_faces = detect_bodies(frame, pose_pairs, map_Idx)
# Initialize MultiTracker
for bbox in face_boxes:
multiTracker.add(cv2.TrackerKCF_create(), frame, tuple(bbox))
# Initialize MultiTracker
for body_box in body_boxes:
multiTracker_body.add(cv2.TrackerKCF_create(), frame, body_box)
compare_res = []
while True:
print("Writing frame {} / {}".format(frame_number + 1, length))
if input_type == "video":
# grab the next frame
(grabbed, frame) = stream.read()
def create_tracker(self): self.tracker_type = "KCF" # Create a tracker tracker = cv2.TrackerKCF_create() return tracker
def tracking():
arduino.write(struct.pack('>B', 0))
global inFrame
global personDetected
global lost_frames
global trackFail
trackFail = False
tracker_types = [
'BOOSTING', 'MIL', 'KCF', 'TLD', 'MEDIANFLOW', 'GOTURN', 'MOSSE',
'CSRT'
]
tracker_type = tracker_types[2]
if int(minor_ver) < 3:
tracker = cv2.Tracker_create(tracker_type)
else:
if tracker_type == 'BOOSTING':
tracker = cv2.TrackerBoosting_create()
if tracker_type == 'MIL':
tracker40 = cv2.TrackerMIL_create()
if tracker_type == 'KCF':
tracker = cv2.TrackerKCF_create()
if tracker_type == 'TLD':
tracker = cv2.TrackerTLD_create()
if tracker_type == 'MEDIANFLOW':
tracker = cv2.TrackerMedianFlow_create()
if tracker_type == 'GOTURN':
tracker = cv2.TrackerGOTURN_create()
if tracker_type == 'MOSSE':
tracker = cv2.TrackerMOSSE_create()
if tracker_type == "CSRT":
tracker = cv2.TrackerCSRT_create()
# Define an initial bounding box
#bbox = (287, 23, 86, 320)
# Uncomment the line below to select a different bounding box
# Initialize tracker with first frame and bounding box
#ok = tracker.init(frame, bbox)
# Why do we have this twice?
#ok, frame = video.read()
while True:
ok, frame = video.read()
while inFrame and personDetected and not SLEEP:
#last_coord = 300, 300
tracker.clear()
tracker = cv2.TrackerKCF_create()
bbox, name = find_face()
#print name + " was detected"
if name != followName:
#lost_frames = lost_frames + 1
#if not lost_frames<3:
if trackFail:
inFrame = False
else:
personDetected = False
continue
trackFail = False
#ask if you are okay because the face cannot be found
#print bbox[2]
ok = tracker.init(frame, bbox)
n = 0
#lost_frames = 0;
while n < 100 and not SLEEP:
# Read a new frame
ok, frame = video.read()
#if not ok:
# break
# Start timertracker = cv2.TrackerKCF_create()
timer = cv2.getTickCount()
# Update tracker
ok, bbox = tracker.update(frame)
x = bbox[0] + bbox[2] / 2
y = bbox[1] + bbox[3] / 2
#print x, y
# Calculate Frames per second (FPS)
fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer)
# Draw bounding box
if ok:
#last_coord = x, y
# Tracking success
#found = True
inFrame = True
#lost_frames = 0
p1 = (int(bbox[0]), int(bbox[1]))
p2 = (int(bbox[0] + bbox[2]), int(bbox[1] + bbox[3]))
cv2.rectangle(frame, p1, p2, (255, 0, 0), 2, 1)
x_center = 320
y_center = 160
x_window = 65
y_window = 20
#print x, y
if not (x_center - x_window) <= x <= (
x_center + x_window
): #Make horizontal displacement prioritized
if x < (x_center - x_window): # face is left
arduino.write(struct.pack('>B', 2))
elif x > (x_center + x_window): # face is right
arduino.write(struct.pack('>B', 3))
else: # face is straight
arduino.write(struct.pack('>B', 0))
else:
if y < (y_center - y_window): # face is too close
arduino.write(struct.pack('>B', 4))
elif y > (y_center + y_window): # face is too far
arduino.write(struct.pack('>B', 1))
else: # face is centered
arduino.write(struct.pack('>B', 0))
else:
# Tracking failure
#lost_frames = lost_frames + 1
cv2.putText(frame, "Tracking failure detected", (100, 80),
cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)
arduino.write(struct.pack('>B', 0))
n = 100
trackFail = True
# Display result
cv2.imshow("Tracking", frame)
# Exit if ESC pressed
k = cv2.waitKey(1) & 0xff
if k == 27: break
n = n + 1
def __init__(self):
self.confThreshold = 0.5 # Confidence threshold (0..1), higher for stricter detection confidence.
self.nmsThreshold = 0.4 # Non-maximum suppression threshold (0..1), higher to remove more duplicate boxes.
self.inpWidth = 160 # Resized image width passed to network
self.inpHeight = 120 # Resized image height passed to network
self.scale = 2 / 255 # Value scaling factor applied to input pixels
self.mean = [127.5, 127.5,
127.5] # Mean BGR value subtracted from input image
self.rgb = True # True if model expects RGB inputs, otherwise it expects BGR
self.bbox = None
self.tracker = cv.TrackerKCF_create()
# Select one of the models:
#model = 'Face' # OpenCV Face Detector, Caffe model
#model = 'MobileNetV2SSD' # MobileNet v2 + SSD trained on Coco (80 object classes), TensorFlow model
#model = 'MobileNetSSD' # MobileNet + SSD trained on Pascal VOC (20 object classes), Caffe model
model = 'MobileNetSSDcoco' # MobileNet + SSD trained on Coco (80 object classes), TensorFlow model
#model = 'YOLOv3' # Darknet Tiny YOLO v3 trained on Coco (80 object classes), Darknet model
#model = 'YOLOv2' # Darknet Tiny YOLO v2 trained on Pascal VOC (20 object classes), Darknet model
# You should not have to edit anything beyond this point.
backend = cv.dnn.DNN_BACKEND_DEFAULT
target = cv.dnn.DNN_TARGET_CPU
self.classes = None
classnames = None
if (model == 'MobileNetSSD'):
classnames = '/jevois/share/darknet/yolo/data/voc.names'
modelname = '/jevois/share/opencv-dnn/detection/MobileNetSSD_deploy.caffemodel'
configname = '/jevois/share/opencv-dnn/detection/MobileNetSSD_deploy.prototxt'
self.rgb = False
elif (model == 'MobileNetV2SSD'):
classnames = '/jevois/share/darknet/yolo/data/coco.names'
modelname = '/jevois/share/opencv-dnn/detection/ssd_mobilenet_v2_coco_2018_03_29.pb'
configname = '/jevois/share/opencv-dnn/detection/ssd_mobilenet_v2_coco_2018_03_29.pbtxt'
elif (model == 'MobileNetSSDcoco'):
classnames = '/jevois/share/darknet/yolo/data/coconew.names'
modelname = '/jevois/share/opencv-dnn/detection/ssd_mobilenet_v1_coco_2017_11_17.pb'
configname = '/jevois/share/opencv-dnn/detection/ssd_mobilenet_v1_coco_2017_11_17.pbtxt'
self.rgb = False
self.nmsThreshold = 0.1
elif (model == 'YOLOv3'):
classnames = '/jevois/share/darknet/yolo/data/coco.names'
modelname = '/jevois/share/darknet/yolo/weights/yolov3-tiny.weights'
configname = '/jevois/share/darknet/yolo/cfg/yolov3-tiny.cfg'
elif (model == 'YOLOv2'):
classnames = '/jevois/share/darknet/yolo/data/voc.names'
modelname = '/jevois/share/darknet/yolo/weights/yolov2-tiny-voc.weights'
configname = '/jevois/share/darknet/yolo/cfg/yolov2-tiny-voc.cfg'
self.inpWidth = 320
self.inpHeight = 240
else:
classnames = '/jevois/share/opencv-dnn/detection/opencv_face_detector.classes'
modelname = '/jevois/share/opencv-dnn/detection/opencv_face_detector.caffemodel'
configname = '/jevois/share/opencv-dnn/detection/opencv_face_detector.prototxt'
self.scale = 1.0
self.mean = [104.0, 177.0, 123.0]
self.rgb = False
# Load names of classes
if classnames:
with open(classnames, 'rt') as f:
self.classes = f.read().rstrip('\n').split('\n')
# Load a network
self.net = cv.dnn.readNet(modelname, configname)
self.net.setPreferableBackend(backend)
self.net.setPreferableTarget(target)
self.timer = jevois.Timer('Neural detection', 10, jevois.LOG_DEBUG)
self.model = model
garbageclasses = [
"shoe", "hat", "eye glasses", "frisbee", "bottle", "plate",
"wine glass", "cup", "fork", "spoon", "bowl", "banana", "apple",
"sandwich", "orange", "broccoli", "carrot", "fruit", "hotdog",
"pizza", "donut", "cake", "vase", "scissors", "toothbrush",
"cardboard", "napkin", "net", "paper", "plastic", "straw"
]
self.garbageclasses = garbageclasses
import cv2
import sys
[major_ver, minor_ver, subminor_ver] = cv2.__version__.split(".")
if int(major_ver) < 3 and int(minor_ver) < 1:
tracker = cv2.TrackerMIL_create()
else:
tracker = cv2.TrackerKCF_create()
# Read video
video = cv2.VideoCapture("Video/CarGame.avi")
#video = cv2.VideoCapture(0)
# Exit if video not opened.
if not video.isOpened():
print("Could not open video")
sys.exit()
# Read first frame.
ok, frame = video.read()
if not ok:
print("Cannot read video file")
sys.exit()
tracker = cv2.MultiTracker_create()
tracker.add(cv2.TrackerKCF_create(), frame, (230, 200, 170, 130))
while cv2.waitKey() != 27:
ok = tracker.add(cv2.TrackerKCF_create(), frame, cv2.selectROI('tracking', frame))
def track_video(video, tracker, bbox, scale, suppress_live_plot, algorithm,
speed):
"""Track a video"""
fps = video.get(cv2.CAP_PROP_FPS)
height, width, _ = FRAME.shape
if not tracker.init(FRAME, bbox):
raise RuntimeError('Could not initialize video file')
frame_number = 0
scaled_bbox = [n / scale for n in bbox]
x_origin = (2.0 * scaled_bbox[0] + scaled_bbox[2]) / 2.0
y_origin = (2.0 * scaled_bbox[1] + scaled_bbox[3]) / 2.0
time_points = [frame_number / fps]
x_points = [x_origin]
y_points = [y_origin]
if not suppress_live_plot:
plt.ion()
fig, axes = plt.subplots(1, 1)
axes.imshow(FRAME)
axes.plot(x_points, y_points)
axes.set_title('Elapsed time: {:d} seconds'.format(int(
time_points[-1])))
axes.set_xlabel(PLOT_XLABEL)
axes.set_ylabel(PLOT_YLABEL)
plt.show()
while True:
try:
for _ in range(speed):
read_frame(video)
frame_number += 1
except RuntimeError:
break
tracking_success, bbox = tracker.update(FRAME)
if not tracking_success:
print(TRACKING_FAIL_MSG)
if algorithm == 'KCF':
tracker = cv2.TrackerKCF_create()
elif algorithm == 'MIL':
tracker = cv2.TrackerMIL_create()
elif algorithm == 'Median-Flow':
tracker = cv2.TrackerMedianFlow_create()
else:
raise ValueError('Unknown algorithm type')
bbox = select_bounding_box()
tracker.init(FRAME, bbox)
scaled_bbox = [n / scale for n in bbox]
#distance = sqrt((((2.0 * scaled_bbox[0] + scaled_bbox[2]) / 2.0) - x_origin)**2
# + (((2.0 * scaled_bbox[1] + scaled_bbox[3]) / 2.0) - y_origin)**2)
time_points.append(frame_number / fps)
x_points.append((2.0 * scaled_bbox[0] + scaled_bbox[2]) / 2.0)
y_points.append((2.0 * scaled_bbox[1] + scaled_bbox[3]) / 2.0)
if not suppress_live_plot:
axes.clear()
axes.imshow(cv2.cvtColor(FRAME, cv2.COLOR_BGR2RGB),
extent=[0, width / scale, height / scale, 0])
axes.plot([
x * (time_points[i] / time_points[-1])
for i, x in enumerate(x_points)
],
y_points,
color='red',
alpha=0.3)
axes.plot(x_points, [
y * (time_points[i] / time_points[-1])
for i, y in enumerate(y_points)
],
color='yellow',
alpha=0.3)
axes.plot([
x * (time_points[i] / time_points[-1])
for i, x in enumerate(x_points)
], [
y * (time_points[i] / time_points[-1])
for i, y in enumerate(y_points)
],
color='green',
alpha=0.5)
axes.plot([
scaled_bbox[0], scaled_bbox[0] + scaled_bbox[2],
scaled_bbox[0] + scaled_bbox[2], scaled_bbox[0], scaled_bbox[0]
], [
scaled_bbox[1], scaled_bbox[1], scaled_bbox[1] +
scaled_bbox[3], scaled_bbox[1] + scaled_bbox[3], scaled_bbox[1]
],
color='white',
alpha=0.3)
axes.set_title('Elapsed time: {:d} seconds'.format(
int(time_points[-1])))
axes.set_xlabel(PLOT_XLABEL)
axes.set_ylabel(PLOT_YLABEL)
axes.set_ylim(height / scale, 0)
plt.pause(0.001)
if not suppress_live_plot:
plt.close()
plt.ioff()
return np.array([time_points, x_points, y_points]).T
