def read_detections(path,
drop_detection_prob: float = 0.0,
add_detection_noise: float = 0.0):
""" parses and converts MOT16 benchmark annotations to known [xmin, ymin, xmax, ymax] format """
path = os.path.expanduser(path)
logger.debug('reading detections from %s' % path)
if not os.path.isfile(path):
raise ValueError('file does not exist')
df = pd.read_csv(path, names=COL_NAMES)
max_frame = df.frame_idx.max()
for frame_idx in range(max_frame):
detections = []
for _, row in df[df.frame_idx == frame_idx].iterrows():
if random.random() < drop_detection_prob:
continue
box = [
row.bb_left, row.bb_top, row.bb_left + row.bb_width,
row.bb_top + row.bb_height
]
if add_detection_noise > 0:
for i in range(4):
box[i] += random.uniform(-add_detection_noise,
add_detection_noise)
detections.append(Detection(box=box))
yield frame_idx, detections
def detectFaceDNN(net, frame, conf_threshold=0.5):
frameOpencvDnn = frame.copy()
frameHeight = frameOpencvDnn.shape[0]
frameWidth = frameOpencvDnn.shape[1]
blob = cv2.dnn.blobFromImage(
frameOpencvDnn,
1.0,
(300, 300),
[104, 117, 123],
False,
False,
)
net.setInput(blob)
detections = net.forward()
bboxes = []
out_detections = []
for i in range(detections.shape[2]):
confidence = detections[0, 0, i, 2]
if confidence > conf_threshold:
x1 = int(detections[0, 0, i, 3] * frameWidth)
y1 = int(detections[0, 0, i, 4] * frameHeight)
x2 = int(detections[0, 0, i, 5] * frameWidth)
y2 = int(detections[0, 0, i, 6] * frameHeight)
bboxes.append([x1, y1, x2, y2])
out_detections.append(
Detection(box=[x1, y1, x2, y2], score=confidence))
cv2.rectangle(
frameOpencvDnn,
(x1, y1),
(x2, y2),
(0, 255, 0),
int(round(frameHeight / 150)),
8,
)
return frameOpencvDnn, bboxes, out_detections
def read_detections(results,
drop_detection_prob: float = 0.0,
add_detection_noise: float = 0.0):
""" parses and converts MOT16 benchmark annotations to known [xmin, ymin, xmax, ymax] format """
detections = []
for i in range(len(results)):
# for _, row in df[df.frame_idx == frame_idx].iterrows():
if random.random() < drop_detection_prob:
continue
box = [
results[i]['xmin'], results[i]['ymin'], results[i]['xmax'],
results[i]['ymax']
]
if add_detection_noise > 0:
for i in range(4):
box[i] += random.uniform(-add_detection_noise,
add_detection_noise)
detections.append(Detection(box=box))
# print('detection box')
# print (box)
# print(Detection(box=box))
return detections
def process_image(self, image: NpImage) -> Sequence[Detection]:
t0 = time.time()
boxes, scores, class_ids = self._predict(image)
elapsed = (time.time() - t0) * 1000.
logger.debug(f'inference time: {elapsed:.3f} ms')
return [
Detection(box=b, score=s, class_id=l)
for b, s, l in zip(boxes, scores, class_ids)
]
def bboxes2out_detections(self, bboxes: BoundingBoxes):
out_detections = []
for bbox in bboxes.bounding_boxes:
out_detections.append(
Detection(box=[bbox.xmin, bbox.ymin, bbox.xmax, bbox.ymax],
score=bbox.probability))
return out_detections
def run():
# prepare multi object tracker
model_spec = {
'order_pos': 1,
'dim_pos': 2,
'order_size': 0,
'dim_size': 2,
'q_var_pos': 5000.,
'r_var_pos': 0.1
}
dt = 1 / 15.0 # assume 15 fps
tracker = MultiObjectTracker(dt=dt, model_spec=model_spec)
# open camera
cap = cv2.VideoCapture(0)
face_detector = FaceDetector()
while True:
ret, frame = cap.read()
if not ret:
break
frame = cv2.resize(frame, dsize=None, fx=0.5, fy=0.5)
# run face detector on current frame
bboxes = face_detector.process(frame)
detections = [Detection(box=bbox) for bbox in bboxes]
logger.debug(f'detections: {detections}')
tracker.step(detections)
tracks = tracker.active_tracks(min_steps_alive=3)
logger.debug(f'tracks: {tracks}')
# preview the boxes on frame
for det in detections:
draw_detection(frame, det)
for track in tracks:
draw_track(frame, track)
cv2.imshow('frame', frame)
# stop demo by pressing 'q'
if cv2.waitKey(int(1000 * dt)) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()
def process_image(self, image: NpImage) -> Sequence[Detection]:
blob = cv2.dnn.blobFromImage(image, 1.0, (300, 300), [104, 117, 123], False, False)
self.net.setInput(blob)
detections = self.net.forward()
# convert output from OpenCV detector to tracker expected format [xmin, ymin, xmax, ymax]
out_detections = []
for i in range(detections.shape[2]):
confidence = detections[0, 0, i, 2]
if confidence > self.conf_threshold:
xmin = int(detections[0, 0, i, 3] * image.shape[1])
ymin = int(detections[0, 0, i, 4] * image.shape[0])
xmax = int(detections[0, 0, i, 5] * image.shape[1])
ymax = int(detections[0, 0, i, 6] * image.shape[0])
out_detections.append(Detection(box=[xmin, ymin, xmax, ymax], score=confidence))
return out_detections
def run(self):
video = self.cam
frame_num = 0
ret, frame = video.read()
height, width = frame.shape[:2]
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
out_video = cv2.VideoWriter('output_23_4.avi', fourcc, 18,
(width, frame_num))
while 1:
# try:
print('--------------------------------')
detection = []
ret, frame = video.read()
frame_num += 1
if frame_num % 1 == 0:
start = time()
detections = self.detector.detect(frame)
for det in detections:
detection.append(Detection(box=np.array(det[:4])))
# draw_detection(frame, Detection(box = np.array(det[:4])))
self.tracker.step(detections=detection)
tracks = self.tracker.active_tracks()
self.process_trackers(frame, tracks)
print("time : ", time() - start)
frame = self.put_res(frame)
frame = cv2.polylines(frame, np.array([self.polygon]), False,
FINAL_LINE_COLOR, 1)
out_video.write(frame)
cv2.imshow('frame', frame)
if cv2.waitKey(10) & 0xFF == ord('q'):
break
# except:
# pass
out_video.release()
cap.release()
cv2.destroyAllWindows()
import numpy as np
from motpy import Detection, MultiObjectTracker
# create a simple bounding box with format of [xmin, ymin, xmax, ymax]
object_box = np.array([1, 1, 10, 10])
# create a multi object tracker with a specified step time of 100ms
tracker = MultiObjectTracker(dt=0.1)
for step in range(10):
# let's simulate object movement by 1 unit (e.g. pixel)
object_box += 1
# update the state of the multi-object-tracker tracker
# with the list of bounding boxes
tracker.step(detections=[Detection(box=object_box)])
# retrieve the active tracks from the tracker (you can customize
# the hyperparameters of tracks filtering by passing extra arguments)
tracks = tracker.active_tracks()
print('MOT tracker tracks %d objects' % len(tracks))
print('first track box: %s' % str(tracks[0].box))
#todo
def run(videoName, dateVar, timeVar):
ageProto = "/root/Project Metro/Models/age_deploy.prototxt"
ageModel = "/root/Project Metro/Models/age_net.caffemodel"
genderProto = "/root/Project Metro/Models/gender_deploy.prototxt"
genderModel = "/root/Project Metro/Models/gender_net.caffemodel"
MODEL_MEAN_VALUES = (78.4263377603, 87.7689143744, 114.895847746)
ageList = [
'(0-3)', '(4-6)', '(8-15)', '(15-18)', '(18-25)', '(30-45)', '(48-55)',
'(60-100)'
]
genderList = ['Male', 'Female']
ageNet = cv2.dnn.readNet(ageModel, ageProto)
genderNet = cv2.dnn.readNet(genderModel, genderProto)
# initialize face detector
#face_detector = cv2.CascadeClassifier("/root/Project Metro/Models/haarcascade_frontalface_default.xml")
net = cv2.dnn.readNetFromCaffe("deploy.prototxt",
"res10_300x300_ssd_iter_140000.caffemodel")
(H, W) = (None, None)
detect_interval = 1
scale_rate = 0.75
show_rate = 1
colours = np.random.rand(32, 3)
c = 0
id_dict = {}
wb = openpyxl.load_workbook('outputgad.xlsx')
ws = wb.worksheets[0]
rowno = ws.max_row + 1
count = ws.max_row - 1
webcam = cv2.VideoCapture(videoName)
fps = webcam.get(cv2.CAP_PROP_FPS)
if fps == 0:
print("No Input Stream Detected")
webcam.release()
cv2.destroyAllWindows()
return
tracker = MultiObjectTracker(
dt=1 / fps,
tracker_kwargs={'max_staleness': 3},
model_spec='constant_acceleration_and_static_box_size_2d',
matching_fn_kwargs={'min_iou': 0.25})
if not webcam.isOpened():
print("No Input Stream Detected")
webcam.release()
cv2.destroyAllWindows()
return
if (videoName == 0):
frameWidth = 500
padding = 25
threshold = 0.5
else:
frameWidth = 800
padding = 20
threshold = 0.8
final_faces = []
while (webcam.isOpened()):
status, frame = webcam.read()
if frame is None:
print("Could not read frame")
webcam.release()
cv2.destroyAllWindows()
break
frame = imutils.resize(frame, width=frameWidth)
(H, W) = frame.shape[:2]
#frame = cv2.resize(frame, (0, 0), fx=scale_rate, fy=scale_rate)
if not status:
print("Could not read frame")
webcam.release()
cv2.destroyAllWindows()
break
#gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
blob = cv2.dnn.blobFromImage(frame, 1.0, (W, H), (104.0, 177.0, 123.0))
net.setInput(blob)
detections = net.forward()
rects = []
for i in range(0, detections.shape[2]):
# filter out weak detections by ensuring the predicted
# probability is greater than a minimum threshold
if detections[0, 0, i, 2] > threshold:
# compute the (x, y)-coordinates of the bounding box for
# the object, then update the bounding box rectangles list
box = detections[0, 0, i, 3:7] * np.array([W, H, W, H])
rects.append(box.astype("int"))
face_list = []
for item in rects:
xmin = item[0]
ymin = item[1]
xmax = item[0] + item[2]
ymax = item[1] + item[3]
face_list.append([xmin, ymin, xmax, ymax])
final_faces = np.array(face_list)
detections = [Detection(box=bbox) for bbox in final_faces]
tracker.step(detections)
tracks = tracker.active_tracks(min_steps_alive=0)
for track in tracks:
d = []
d = track.box
d_id = track.id
d = d.astype(np.int32)
x = d[0]
y = d[1]
w = d[2]
h = d[3]
if d_id not in id_dict.keys():
#face = frame[y:h,x:w]
face = frame[max(0, y -
padding):min(h + padding, frame.shape[0] - 1),
max(0, x -
padding):min(w + padding, frame.shape[1] - 1)]
blob = cv2.dnn.blobFromImage(face,
1.0, (227, 227),
MODEL_MEAN_VALUES,
swapRB=False)
genderNet.setInput(blob)
genderPreds = genderNet.forward()
gender = genderList[genderPreds[0].argmax()]
print('Person :', count + 1)
print(f'Gender: {gender}')
ageNet.setInput(blob)
agePreds = ageNet.forward()
age = ageList[agePreds[0].argmax()]
print(f'Age: {age[1:-1]} years')
id_dict[d_id] = gender
c1 = ws.cell(row=rowno, column=1)
c2 = ws.cell(row=rowno, column=2)
c3 = ws.cell(row=rowno, column=3)
c4 = ws.cell(row=rowno, column=4)
c1.value = dateVar
c2.value = timeVar
c3.value = gender
c4.value = age[1:-1]
wb.save('outputgad.xlsx')
count += 1
rowno += 1
cv2.imwrite(
"{0}/{1}{2}_{3}.jpg".format("facepics", gender, age, d_id),
face)
cv2.rectangle(frame, (x, y), (w - x, h - y), (0, 255, 0), 2)
cv2.putText(frame, f'{gender}, {age}', (d[0], d[1] - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 255), 2,
cv2.LINE_AA)
resultframe = cv2.resize(frame, (0, 0), fx=show_rate, fy=show_rate)
cv2.imshow("Detecting age and gender", resultframe)
if cv2.waitKey(1) & 0XFF == ord('q'):
print("Detection Stopped Manually")
webcam.release()
cv2.destroyAllWindows()
break
def detect(save_img=False):
source, weights, view_img, save_txt, imgsz = opt.source, opt.weights, opt.view_img, opt.save_txt, opt.img_size
webcam = source.isnumeric() or source.endswith(
'.txt') or source.lower().startswith(('rtsp://', 'rtmp://', 'http://'))
# Directories
save_dir = Path(
increment_path(Path(opt.project) / opt.name,
exist_ok=opt.exist_ok)) # increment run
(save_dir / 'labels' if save_txt else save_dir).mkdir(
parents=True, exist_ok=True) # make dir
# Initialize
set_logging()
device = select_device(opt.device)
half = device.type != 'cpu' # half precision only supported on CUDA
#create a multi object tracker
tracker = MultiObjectTracker(dt=0.1)
# Load model
model = attempt_load(weights, map_location=device) # load FP32 model
stride = int(model.stride.max()) # model stride
imgsz = check_img_size(imgsz, s=stride) # check img_size
if half:
model.half() # to FP16
# Second-stage classifier
classify = False
if classify:
modelc = load_classifier(name='resnet101', n=2) # initialize
modelc.load_state_dict(
torch.load('weights/resnet101.pt',
map_location=device)['model']).to(device).eval()
# Set Dataloader
vid_path, vid_writer = None, None
if webcam:
view_img = check_imshow()
cudnn.benchmark = True # set True to speed up constant image size inference
dataset = LoadStreams(source, img_size=imgsz, stride=stride)
else:
save_img = True
dataset = LoadImages(source, img_size=imgsz, stride=stride)
# Get names and colors
names = model.module.names if hasattr(model, 'module') else model.names
colors = [[random.randint(0, 255) for _ in range(3)] for _ in names]
# Run inference
if device.type != 'cpu':
model(
torch.zeros(1, 3, imgsz, imgsz).to(device).type_as(
next(model.parameters()))) # run once
t0 = time.time()
for path, img, im0s, vid_cap in dataset:
img = torch.from_numpy(img).to(device)
img = img.half() if half else img.float() # uint8 to fp16/32
img /= 255.0 # 0 - 255 to 0.0 - 1.0
if img.ndimension() == 3:
img = img.unsqueeze(0)
# Inference
t1 = time_synchronized()
pred = model(img, augment=opt.augment)[0]
# Apply NMS
pred = non_max_suppression(pred,
opt.conf_thres,
opt.iou_thres,
classes=opt.classes,
agnostic=opt.agnostic_nms)
t2 = time_synchronized()
# Apply Classifier
if classify:
pred = apply_classifier(pred, modelc, img, im0s)
# Process detections
for i, det in enumerate(pred): # detections per image
if webcam: # batch_size >= 1
p, s, im0, frame = path[i], '%g: ' % i, im0s[i].copy(
), dataset.count
else:
p, s, im0, frame = path, '', im0s, getattr(dataset, 'frame', 0)
p = Path(p) # to Path
save_path = str(save_dir / p.name) # img.jpg
txt_path = str(save_dir / 'labels' / p.stem) + (
'' if dataset.mode == 'image' else f'_{frame}') # img.txt
s += '%gx%g ' % img.shape[2:] # print string
gn = torch.tensor(im0.shape)[[1, 0, 1,
0]] # normalization gain whwh
if len(det):
# Rescale boxes from img_size to im0 size
det[:, :4] = scale_coords(img.shape[2:], det[:, :4],
im0.shape).round()
# Print results
for c in det[:, -1].unique():
n = (det[:, -1] == c).sum() # detections per class
s += f"{n} {names[int(c)]}{'s' * (n > 1)}, " # add to string
# Write results
out_detections = []
for *xyxy, conf, cls in reversed(det):
object_box = np.array([
int(xyxy[0]),
int(xyxy[1]),
int(xyxy[2]),
int(xyxy[3])
])
out_detections.append(
Detection(box=object_box, score=conf.to('cpu')))
tracker.step(out_detections)
tracks = tracker.active_tracks(3)
for track in tracks:
label = f'{track.id[:5]}'
plot_one_box(track.box,
im0,
label=label,
color=colors[int(cls)],
line_thickness=3)
# Print time (inference + NMS)
print(f'{s}Done. ({t2 - t1:.3f}s)')
# Stream results
if view_img:
cv2.imshow(str(p), im0)
key = cv2.waitKey(1) # 1 millisecond
if key == ord('q'):
break
print(f'Done. ({time.time() - t0:.3f}s)')
def run(videoName, dateVar, timeVar):
ageProto = "/root/Project Metro/Models/age_deploy.prototxt"
ageModel = "/root/Project Metro/Models/age_net.caffemodel"
genderProto = "/root/Project Metro/Models/gender_deploy.prototxt"
genderModel = "/root/Project Metro/Models/gender_net.caffemodel"
MODEL_MEAN_VALUES = (78.4263377603, 87.7689143744, 114.895847746)
ageList = [
'(0-3)', '(4-6)', '(8-15)', '(15-18)', '(18-25)', '(30-45)', '(48-55)',
'(60-100)'
]
genderList = ['Male', 'Female']
ageNet = cv2.dnn.readNet(ageModel, ageProto)
genderNet = cv2.dnn.readNet(genderModel, genderProto)
# initialize face detector
face_detector = cv2.CascadeClassifier(
"/root/Project Metro/Models/haarcascade_frontalface_default.xml")
detect_interval = 1
scale_rate = 0.75
show_rate = 1
colours = np.random.rand(32, 3)
tracker = MultiObjectTracker(
dt=1 / 25,
tracker_kwargs={'max_staleness': 3},
model_spec='constant_acceleration_and_static_box_size_2d',
matching_fn_kwargs={'min_iou': 0.25})
c = 0
id_dict = {}
wb = openpyxl.load_workbook('outputgad.xlsx')
ws = wb.worksheets[0]
rowno = ws.max_row + 1
count = ws.max_row - 1
webcam = cv2.VideoCapture(videoName)
fps = webcam.get(cv2.CAP_PROP_FPS)
if fps == 0:
print("No Input Stream Detected")
webcam.release()
cv2.destroyAllWindows()
return
tracker = MultiObjectTracker(
dt=1 / fps,
tracker_kwargs={'max_staleness': 3},
model_spec='constant_acceleration_and_static_box_size_2d',
matching_fn_kwargs={'min_iou': 0.25})
if not webcam.isOpened():
print("No Input Stream Detected")
webcam.release()
cv2.destroyAllWindows()
return
padding = 20
final_faces = []
while (webcam.isOpened()):
status, frame = webcam.read()
#frame = cv2.resize(frame, (0, 0), fx=scale_rate, fy=scale_rate)
if not status:
print("Could not read frame")
webcam.release()
cv2.destroyAllWindows()
return
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
if c % detect_interval == 0:
#faces = face_detector.detectMultiScale(gray, 1.3, 5)
faces = face_detector.detectMultiScale(gray,
scaleFactor=2,
minNeighbors=5,
minSize=(300, 300))
faces = np.array(faces)
face_sums = faces.shape[0]
if face_sums > 0:
face_list = []
for item in faces:
xmin = item[0]
ymin = item[1]
xmax = item[0] + item[2]
ymax = item[1] + item[3]
face_list.append([xmin, ymin, xmax, ymax])
final_faces = np.array(face_list)
detections = [Detection(box=bbox) for bbox in final_faces]
tracker.step(detections)
tracks = tracker.active_tracks(min_steps_alive=0)
for track in tracks:
d = []
d = track.box
d_id = track.id
d = d.astype(np.int32)
x = d[0]
y = d[1]
w = d[2]
h = d[3]
if d_id not in id_dict.keys():
face = frame[max(0, y -
padding):min(h + padding, frame.shape[0] - 1),
max(0, x -
padding):min(w + padding, frame.shape[1] - 1)]
blob = cv2.dnn.blobFromImage(face,
1.0, (227, 227),
MODEL_MEAN_VALUES,
swapRB=False)
genderNet.setInput(blob)
genderPreds = genderNet.forward()
gender = genderList[genderPreds[0].argmax()]
print('Person :', count + 1)
print(f'Gender: {gender}')
ageNet.setInput(blob)
agePreds = ageNet.forward()
age = ageList[agePreds[0].argmax()]
print(f'Age: {age[1:-1]} years')
id_dict[d_id] = gender
c1 = ws.cell(row=rowno, column=1)
c2 = ws.cell(row=rowno, column=2)
c3 = ws.cell(row=rowno, column=3)
c4 = ws.cell(row=rowno, column=4)
c1.value = dateVar
c2.value = timeVar
c3.value = gender
c4.value = age[1:-1]
wb.save('outputgad.xlsx')
count += 1
rowno += 1
cv2.imwrite(
"{0}/{1}{2}_{3}.jpg".format("facepics", gender, age, d_id),
face)
cv2.rectangle(frame, (x, y), (w, h), (0, 255, 0), 2)
cv2.putText(frame, f'{gender}, {age}', (d[0], d[1] - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 255), 2,
cv2.LINE_AA)
resultframe = cv2.resize(frame, (0, 0), fx=show_rate, fy=show_rate)
cv2.imshow("Detecting age and gender", resultframe)
if cv2.waitKey(1) & 0XFF == ord('q'):
print("Detection Stopped Manually")
webcam.release()
cv2.destroyAllWindows()
return
break
def run():
# prepare multi object tracker
model_spec = {
'order_pos': 1,
'dim_pos': 2,
'order_size': 0,
'dim_size': 2,
'q_var_pos': 5000.,
'r_var_pos': 0.1
}
# model_spec = {
# 'order_pos': 1, 'dim_pos': 2, # position is a center in 2D space; under constant velocity model
# 'order_size': 0, 'dim_size': 2, # bounding box is 2 dimensional; under constant velocity model
# 'q_var_pos': 1000., # process noise
# 'r_var_pos': 0.1 # measurement noise
# }
# tracker = MultiObjectTracker(dt=1 / 10, model_spec=model_spec)
dt = 1 / 15.0 # assume 8 fps
tracker = MultiObjectTracker(dt=dt, model_spec=model_spec)
input_video = args.input_video
# open camera
cap = cv2.VideoCapture(input_video)
# vid = imageio.get_reader(input_video, 'ffmpeg')
people_detector = PeopleDetector()
while (True):
ret, frame = cap.read()
# frame = cv2.resize(frame, dsize=None, fx=0.5, fy=0.5)
# run face detector on current frame
bboxes = people_detector.process(frame, args.confidence)
detections = [Detection(box=bbox) for bbox in bboxes]
logger.debug(f'detections: {detections}')
tracker.step(detections)
tracks = tracker.active_tracks(min_steps_alive=3)
logger.debug(f'tracks: {tracks}')
# preview the boxes on frame
for det in detections:
draw_detection(frame, det)
for track in tracks:
draw_track(frame, track)
if cv2.waitKey(1) & 0xFF == ord('q') or ret == False:
cap.release()
cv2.destroyAllWindows()
break
cv2.imshow('frame', frame)
# stop demo by pressing 'q'
if cv2.waitKey(int(1000 * dt)) & 0xFF == ord('q'):
break
# cap.release()
cv2.destroyAllWindows()
def main():
ID_only = []
verbose = args.verbose
if args.model == 'yolov3':
CONFIG_PATH, WEIGHTS_PATH = 'yolov3.cfg', 'yolov3.weights'
if not os.path.isfile(WEIGHTS_PATH):
logger.debug('downloading model...')
urlretrieve('https://pjreddie.com/media/files/yolov3.weights', WEIGHTS_PATH)
if args.input_video == 'mall':
input_video = 'sample_mall_vid.mp4'
fx, fy = 1, 1
x1_loc = 140
y1_loc = 240
x2_loc = 340
y2_loc = 250
elif args.input_video == 'shop':
input_video = 'sample_shop_vid.mp4'
fx, fy = 0.7, 0.7
x1_loc = 600
y1_loc = 500
x2_loc = 740
y2_loc = 390
update_text_font = ImageFont.truetype("arial.ttf", 15)
# Load names of classes and get random colors
classes = open('coco.names').read().strip().split('\n')
accepted_classes = ['person', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase']
accessory_ref_lst = ['backpack', 'umbrella', 'handbag', 'tie', 'suitcase']
inner_keys = ['object', 'time', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase']
idx_accepted = [0, 24, 25, 26, 27, 28]
np.random.seed(42)
colors = np.random.randint(0, 255, size=(len(classes), 3), dtype='uint8')
# Give the configuration and weight files for the model and load the network.
net = cv.dnn.readNetFromDarknet('yolov3.cfg', 'yolov3.weights')
net.setPreferableBackend(cv.dnn.DNN_BACKEND_OPENCV)
# net.setPreferableTarget(cv.dnn.DNN_TARGET_CPU)
if verbose: print('model loaded')
# determine the output layer
ln = net.getLayerNames()
ln = [ln[i[0] - 1] for i in net.getUnconnectedOutLayers()]
# open camera
cap = cv2.VideoCapture(input_video)
dt = 1 / 8.0 # assume 8 fps
# prepare multi object tracker
model_spec = {'order_pos': 1, 'dim_pos': 2,
'order_size': 0, 'dim_size': 2,
'q_var_pos': 5000., 'r_var_pos': 0.1}
# prepare tracking
tracker = MultiObjectTracker(dt=dt, model_spec=model_spec)
# python dictionary to track people
d = {
'ID':
{'object': 'value_1', 'time': 'value_2', 'backpack': 'value_3', 'umbrella': 'value_4', 'handbag': 'value_5', 'tie': 'value_6', 'suitcase': 'value_7'}
}
d_bbox = {'ID':
{'x1': 'value_1', 'y1': 'value_2', 'x2': 'value_3', 'y2': 'value_4'}
}
arr_d = []
ctr = 0
clear_ctr = 0
img_array = []
while(True):
# only process every 30 frames
if args.input_video == 'shop' and ctr < 45:
# shop example frozen for the first 40 frames
ret, img = cap.read()
ctr += 1
continue
# while True:
ret, img = cap.read()
clear_ctr += 1
# save if end of video file
if img is None:
if args.save_bool:
save_video(args.input_video, img_array, size)
# exit
break
img = cv2.resize(img, dsize=None, fx=fx, fy=fy)
size = (img.shape[1], img.shape[0])
# construct a blob from the image
blob = cv.dnn.blobFromImage(img, 1/255.0, (416, 416), swapRB=True, crop=False)
net.setInput(blob)
if verbose: t0 = time.time()
outputs = net.forward(ln)
if verbose: t = time.time()
if verbose: print('time=', t-t0)
boxes = []
confidences = []
classIDs = []
h, w = img.shape[:2]
for output in outputs:
for detection in output:
scores = detection[5:]
classID = np.argmax(scores)
# ignore if not in classes we want
if classID not in idx_accepted:
continue
# logger.debug(f'class: {classes[classID]}')
confidence = scores[classID]
if confidence > 0.5:
box = detection[:4] * np.array([w, h, w, h])
(centerX, centerY, width, height) = box.astype("int")
x = int(centerX - (width / 2))
y = int(centerY - (height / 2))
box = [x, y, int(width), int(height)]
boxes.append(box)
confidences.append(float(confidence))
classIDs.append(classID)
indices = cv.dnn.NMSBoxes(boxes, confidences, 0.5, 0.4)
class_lst = []
bboxes = []
if len(indices) > 0:
for i in indices.flatten():
(x, y) = (boxes[i][0], boxes[i][1])
(w, h) = (boxes[i][2], boxes[i][3])
# # old version of boxes without ID tracking
# color = [int(c) for c in colors[classIDs[i]]]
# cv.rectangle(img, (x, y), (x + w, y + h), color, 2)
# text = "{}: {:.4f}".format(classes[classIDs[i]], confidences[i])
# cv.putText(img, text, (x, y - 5), cv.FONT_HERSHEY_SIMPLEX, 0.5, color, 1)
class_lst.append(classes[classIDs[i]])
# getting the boundaries of the box for tracking
xmin = int(x)
ymin = int(y)
xmax = int(x + w)
ymax = int(y + h)
bboxes.append([xmin, ymin, xmax, ymax])
# if empty list
if not class_lst:
continue
''' detection adapated from https://learnopencv.com/goturn-deep-learning-based-object-tracking/ '''
detections = [Detection(box=bbox) for bbox in bboxes]
if verbose: logger.debug(f'detections: {detections}')
# edited MOTPY tracker source code
tracker.step(detections, class_lst)
tracks = tracker.active_tracks(min_steps_alive=-1)
if verbose: logger.debug(f'tracks: {tracks}')
# prepare text for each person detected
# text_arr = []
# # preview the boxes on frame
# for det in detections:
# draw_detection(img, det)
u_x_p = []
u_y_p = []
u_x_a = []
u_y_a = []
people_track_lst = []
accessories_track_lst = []
for idx, track in enumerate(tracks):
bound_box = track[1]
ID = track[0].split('-')[0]
class_ID = track[0].split('-')[1]
# append to sort
if class_ID == 'person':
people_track_lst.append(track)
u_x_p.append(mean([bound_box[0], bound_box[2]]))
u_y_p.append(mean([bound_box[1], bound_box[3]]))
custom_draw_track(img, track, 'person')
else:
accessories_track_lst.append(track)
u_x_a.append(mean([bound_box[0], bound_box[2]]))
u_y_a.append(mean([bound_box[1], bound_box[3]]))
custom_draw_track(img, track, 'accessory')
# custom_draw_track(img, track, text_arr[idx])
time_stamp = time.strftime("%Y%m%d%H%M%S")
# combine the track list, but accessories ordered last
track_list = people_track_lst + accessories_track_lst
ux = u_x_p + u_x_a
uy = u_y_p + u_y_a
# determine how many people detected
if len(indices) > 0:
# process bag and count people
for idx, track in enumerate(track_list):
bound_box = track[1]
ID = track[0].split('-')[0]
class_ID = track[0].split('-')[1]
bp_curr = None
ub_curr = None
hb_curr = None
t_curr = None
sc_curr = None
status_stamp = None
px_h = None
# if accessory
if class_ID != 'person':
# calculate a list of distances between the people and this point
person_index, img = distance_2d(ux[idx], uy[idx], u_x_p, u_y_p, img)
# if it was not registered as an accessory yet
# if exists(ID, arr_d) is False:
# Check if key exist in dictionary using any()
if any(ID in d_t.values() for d_t in d.values()) is False:
# index of the person...
curr_person = people_track_lst[person_index]
owner_ID = curr_person[0].split('-')[0]
# set the new value into the dictionary
d[owner_ID][class_ID] = ID
# add to dictionary (changed to list) if it doesn't exist
# elif exists(ID, arr_d) is False:
elif ID not in d.keys():
d.update({
ID: {'object': class_ID, 'time': time_stamp, 'status': status_stamp, 'height': px_h, 'backpack': bp_curr, 'umbrella': ub_curr, 'handbag': hb_curr, 'tie': t_curr, 'suitcase': sc_curr}
})
d_bbox.update({
ID: {'x1': [round(bound_box[0])], 'y1': [round(bound_box[1])], 'x2': [round(bound_box[2])], 'y2': [round(bound_box[3])]}
})
# every two frames, we append
elif clear_ctr % 2:
# it's already in the list, we append the position
d_bbox[ID]['x1'].append(round(bound_box[0]))
d_bbox[ID]['y1'].append(round(bound_box[1]))
d_bbox[ID]['x2'].append(round(bound_box[2]))
d_bbox[ID]['y2'].append(round(bound_box[3]))
# arr_d.append([ID, class_ID, time_stamp, bp_curr, ub_curr, hb_curr, t_curr, sc_curr])
# ID_only.append(ID)
# print(d_bbox)
# every 20 frames, we remove idle status objects from the dictionaries
if clear_ctr % 2:
d, d_bbox = clean_bbox_dict(d, d_bbox)
# print(d)
# print(ID_only)
num_people = len(people_track_lst)
# get time stamp
img = write_stats(img, num_people, time_stamp, update_text_font)
if verbose: logger.debug(f'number of people: {num_people}, time of day: {time_stamp}')
# draw line for people counting
img = draw_line(img, x1_loc, y1_loc, x2_loc, y2_loc, (0, 0, 255), 5)
cv.imshow('window', img)
# stop demo by pressing 'q'
if cv2.waitKey(int(1000*dt)) & 0xFF == ord('q'):
break
img_array.append(img)
if args.SlowMode:
input("Press Enter to continue...")
with open('shop.json', 'w') as json_file:
json.dump(d, json_file, indent=4)
# uncomment to route!
if args.flask_bool: return Response(response=str(d), status=200,mimetype="application/json")
def track_hoa_df(
hoa_dets,
dt=0.02,
start_frame=0,
end_frame=100,
video_id=None,
verbose=True,
object_only=False,
keep_longest=True,
):
"""
Args:
keep_longest (bool): find longest object track sequence
"""
# Initialize track lists and tracker
obj_tracker = MultiObjectTracker(dt=dt)
tracked_obj = []
if not object_only:
lh_tracker = MultiObjectTracker(dt=dt)
rh_tracker = MultiObjectTracker(dt=dt)
# Intialize tracked dicts
tracked_lh = []
tracked_rh = []
# Last non-empty df
for frame_idx in tqdm(range(start_frame, end_frame)):
hoa_df = hoa_dets[hoa_dets.frame == frame_idx]
obj_df = hoa_df[hoa_df.det_type == "object"]
obj_dets = [
Detection(gethoa.row2box(row)) for _, row in obj_df.iterrows()
]
obj_tracker.step(detections=obj_dets)
tracked_obj.extend(
trackconv.track2dicts(
obj_tracker.active_tracks(),
frame_idx,
video_id=video_id,
det_type="object",
)
)
if not object_only:
lh_df = hoa_df[
(hoa_df.det_type == "hand") & (hoa_df.side == "left")
]
rh_df = hoa_df[
(hoa_df.det_type == "hand") & (hoa_df.side == "right")
]
lh_dets = [
Detection(gethoa.row2box(row)) for _, row in lh_df.iterrows()
]
rh_dets = [
Detection(gethoa.row2box(row)) for _, row in rh_df.iterrows()
]
lh_tracker.step(detections=lh_dets)
rh_tracker.step(detections=rh_dets)
tracked_lh.extend(
trackconv.track2dicts(
lh_tracker.active_tracks(),
frame_idx,
video_id=video_id,
det_type="hand",
side="left",
)
)
tracked_rh.extend(
trackconv.track2dicts(
rh_tracker.active_tracks(),
frame_idx,
video_id=video_id,
det_type="hand",
side="right",
)
)
if verbose:
obj_tracks = pd.DataFrame(tracked_obj)
if keep_longest:
longest_track_idx = (
obj_tracks.groupby("track_id").frame.nunique().idxmax()
)
# Filter object which has longest track
tracked_obj = obj_tracks[obj_tracks.track_id == longest_track_idx]
print_track_info(tracked_obj)
if not object_only:
lh_tracks = pd.DataFrame(tracked_lh)
rh_tracks = pd.DataFrame(tracked_rh)
print_track_info(lh_tracks, track_type="left hand")
print_track_info(rh_tracks, track_type="right hand")
tracked_hoa = pd.DataFrame(
tracked_obj.to_dict("records") + tracked_lh + tracked_rh
)
else:
tracked_hoa = pd.DataFrame(tracked_obj)
if keep_longest:
start_track_frame = tracked_obj.frame.min()
end_track_frame = tracked_obj.frame.max()
# Keep only region that focuses on longest track
tracked_hoa = tracked_hoa[
(tracked_hoa.frame >= start_track_frame)
& (tracked_hoa.frame <= end_track_frame)
]
return tracked_hoa
