def track_viou_matlab_wrapper(
frames_path,
detections,
sigma_l,
sigma_h,
sigma_iou,
t_min,
ttl,
tracker_type,
keep_upper_height_ratio=1.0,
):
"""
Matlab wrapper of the v-iou tracker for the detrac evaluation toolkit.
Args:
detections (numpy.array): numpy array of detections, usually supplied by run_tracker.m
sigma_l (float): low detection threshold.
sigma_h (float): high detection threshold.
sigma_iou (float): IOU threshold.
t_min (float): minimum track length in frames.
Returns:
float: speed in frames per second.
list: list of tracks.
"""
detections = detections.reshape((7, -1)).transpose()
dets = load_mot(detections, with_classes=False)
start = time()
tracks = track_viou(
frames_path + "img{:05d}.jpg",
dets,
sigma_l,
sigma_h,
sigma_iou,
int(t_min),
int(ttl),
tracker_type,
keep_upper_height_ratio,
)
end = time()
id_ = 1
out = []
for track in tracks:
for i, bbox in enumerate(track["bboxes"]):
out += [
float(bbox[0]),
float(bbox[1]),
float(bbox[2] - bbox[0]),
float(bbox[3] - bbox[1]),
float(track["start_frame"] + i),
float(id_),
]
id_ += 1
num_frames = len(dets)
speed = num_frames / (end - start)
return speed, out
def track_iou_matlab_wrapper(detections, sigma_l, sigma_h, sigma_iou, t_min):
"""
Matlab wrapper of the iou tracker for the detrac evaluation toolkit.
Args:
detections (numpy.array): numpy array of detections, usually supplied by run_tracker.m
sigma_l (float): low detection threshold.
sigma_h (float): high detection threshold.
sigma_iou (float): IOU threshold.
t_min (float): minimum track length in frames.
Returns:
float: speed in frames per second.
list: list of tracks.
"""
detections = detections.reshape((7, -1)).transpose()
dets = load_mot(detections)
start = time()
tracks = track_iou(dets, sigma_l, sigma_h, sigma_iou, t_min)
end = time()
id_ = 1
out = []
for track in tracks:
for i, bbox in enumerate(track['bboxes']):
out += [float(bbox[0]), float(bbox[1]), float(bbox[2] - bbox[0]), float(bbox[3] - bbox[1]),
float(track['start_frame'] + i), float(id_)]
id_ += 1
num_frames = len(dets)
speed = num_frames / (end - start)
return speed, out
def info_callback(self, info):
bbox = np.empty((0,5), float) #bbox라는 한 객체를 담는 객체 생성
# 모든 message data를 객체마다 bbox에 저장
for i in range( len(info.centerX) ):
bbox = np.append(bbox, [[info.box_left[i], info.box_top[i], info.box_right[i], info.box_bottom[i], info.conf[i]]], axis=0)
detection = load_mot(bbox) # bbox로부터 데이터 읽은 후 id부여해서 return
self.pre_tracker, self.img_tracks = track_iou( self.pre_tracker, detection)
def tracker(args, detections):
'''runs a detection, object-oriented version of demo.py'''
#might need to add nms to load_mot
dets = load_mot(detections, with_classes=0)
print('\t[-] Track with Intersection over Union')
tracks = track_iou(dets, args.sigma_l, args.sigma_h, args.sigma_iou,
args.t_min)
print('\t[-] Format Tracks')
tracks = build_array2(tracks, fmt=args.fmt)
return tracks
def main(args):
detections = load_mot(args.detection_path)
start = time()
tracks = track_iou(detections, args.sigma_l, args.sigma_h, args.sigma_iou, args.t_min)
end = time()
num_frames = len(detections)
print("finished at " + str(int(num_frames / (end - start))) + " fps!")
save_to_csv(args.output_path, tracks)
def main(args):
detections = load_mot(args.detection_path)
start = time()
tracks = track_iou(detections, args.sigma_l, args.sigma_h, args.sigma_iou,
args.t_min)
end = time()
num_frames = len(detections)
print("finished at " + str(int(num_frames / (end - start))) + " fps!")
save_to_csv(args.output_path, tracks)
def main(args):
formats = ["motchallenge", "visdrone"]
assert (
args.format
in formats), "format '{}' unknown supported formats are: {}".format(
args.format, formats)
with_classes = False
if args.format == "visdrone":
with_classes = True
detections = load_mot(
args.detection_path,
nms_overlap_thresh=args.nms,
with_classes=with_classes,
min_short=args.min_short,
max_short=args.max_short,
max_long=args.max_long,
)
if args.visual:
tracks = track_viou(
args.frames_path,
detections,
args.sigma_l,
args.sigma_h,
args.sigma_iou,
args.t_min,
args.ttl,
args.visual,
args.keep_upper_height_ratio,
)
else:
if with_classes:
# track_viou can also be used without visual tracking, but note that the speed will be much slower compared
# to track_iou. However, this way supports the optimal LAP solving and the handling of multiple object classes:
tracks = track_viou(
args.frames_path,
detections,
args.sigma_l,
args.sigma_h,
args.sigma_iou,
args.t_min,
args.ttl,
"NONE",
args.keep_upper_height_ratio,
)
else:
tracks = track_iou(detections, args.sigma_l, args.sigma_h,
args.sigma_iou, args.t_min)
save_to_csv(args.output_path, tracks, fmt=args.format)
def main(datasetPath):
if 'Fluo-N2DL-HeLa' in datasetPath:
mode = 'Fluo-N2DL-HeLa'
elif 'PhC-C2DL-PSC' in datasetPath:
mode = 'PhC-C2DL-PSC'
else:
mode = 'DIC-C2DH-HeLa'
detPath = datasetPath.replace(r'/', '_') + '.txt'
detAbsPath = os.path.join('det', detPath)
genDets(datasetPath, mode)
print('loading detections...')
detections = load_mot(detAbsPath,
nms_overlap_thresh=cfg.nms,
with_classes=False)
print('tracking...')
tracks = track_iou(detections, cfg.sigma_l, cfg.sigma_h, cfg.sigma_iou,
cfg.t_min, cfg.max_missing)
save_to_csv(os.path.join(cfg.output_path, detPath), tracks, fmt=cfg.format)
def main(args):
with open(args.seqmap) as fd:
seqs = [line.rstrip('\n') for line in fd]
for idx, seq in enumerate(seqs):
if seq == "name" or seq == "":
continue
else:
if "DPM" in seq:
sigma_l = -0.5
sigma_h = 0.5
sigma_iou = 0.4
t_min = 4
elif "FRCNN" in seq:
sigma_l = 0.0
sigma_h = 0.9
sigma_iou = 0.3
t_min = 3
elif "SDP" in seq:
sigma_l = 0.4
sigma_h = 0.5
sigma_iou = 0.2
t_min = 2
else:
print("No detector name found, this could happen with the "
"wrong seqmap seqmap file. "
"Please use c10-train.txt, c10-test.txt or c10-all.txt")
exit()
# Take the pre-computed detections.
det_path = args.benchmark_dir + "/" + seq + "/det/det.txt"
out_path = args.res_dir + "/" + seq + ".txt"
detections = load_mot(det_path)
start = time()
tracks = track_iou(detections, sigma_l, sigma_h, sigma_iou, t_min)
end = time()
num_frames = len(detections)
print("finished " + seq + " at " +
str(int(num_frames / (end - start))) + " fps!")
save_to_csv(out_path, tracks)
def callback(self, image, info):
try:
cv_image = self.bridge.imgmsg_to_cv2(image, "bgr8")
except CvBridgeError as e:
print(e)
bbox = np.empty((0, 5), float)
for i in range(len(info.centerX)):
bbox = np.append(bbox, [[
info.box_left[i], info.box_top[i], info.box_right[i],
info.box_bottom[i], info.conf[i], info.distance[i]
]],
axis=0)
detection = load_mot(bbox)
# print("detection")
# print(detection)
# for i in range(len(detection)):
# print("detection["+str(i)+"]")
# print(detection[i])
self.pre_tracker, tracks = track_iou(self.pre_tracker, detection, 0.3)
if tracks is "wait":
print("wait")
elif tracks is None and self.get_age() >= 5:
print("실패")
elif tracks is None and self.get_age() < 5:
self.set_age(1)
print("aging")
else:
self.set_age(0)
print("성공")
blue_color = (255, 0, 0)
cv.rectangle(cv_image,
(int(tracks['bbox'][0]), int(tracks['bbox'][1])),
(int(tracks['bbox'][2]), int(tracks['bbox'][3])),
blue_color, 3)
control_msg = self.bridge.cv2_to_imgmsg(cv_image, "bgr8")
self.control_pub.publish(control_msg)
def main(args):
with open(args.seqmap) as fd:
seqs = [line.rstrip('\n') for line in fd]
for idx, seq in enumerate(seqs):
if seq == "name" or seq == "":
continue
else:
if "DPM" in seq:
sigma_l = -0.5
sigma_h = 0.5
sigma_iou = 0.4
t_min = 4
elif "FRCNN" in seq:
sigma_l = 0.0
sigma_h = 0.9
sigma_iou = 0.3
t_min = 3
elif "SDP" in seq:
sigma_l = 0.4
sigma_h = 0.5
sigma_iou = 0.2
t_min = 2
else:
print("No detector name found, this could happen with the wrong seqmap seqmap file. "
"Please use c10-train.txt, c10-test.txt or c10-all.txt")
exit()
det_path = args.benchmark_dir + "/" + seq + "/det/det.txt"
out_path = args.res_dir + "/" + seq + ".txt"
detections = load_mot(det_path)
start = time()
tracks = track_iou(detections, sigma_l, sigma_h, sigma_iou, t_min)
end = time()
num_frames = len(detections)
print("finished " + seq + " at " + str(int(num_frames / (end - start))) + " fps!")
save_to_csv(out_path, tracks)
def main(args):
print('loading detections...')
det_path = os.path.expanduser('~/') + args.detection_path
detections = load_mot(det_path)
print('tracking...')
start = time()
tracks = track_iou(detections, args.sigma_l, args.sigma_h, args.sigma_iou,
args.t_min)
end = time()
print("Finished Tracking", tracks)
num_frames = len(detections)
print("finished at " + str(int(num_frames / (end - start))) + " fps!")
orig_out_path = os.path.expanduser('~/') + args.output_path
save_to_csv(orig_out_path, tracks)
out_path = os.path.expanduser('~/') + args.output_path + 'ordredfrm'
if args.frameorder:
#reading in all the data again for re-formatting
idorder_frameorder(orig_out_path, out_path)
def track_iou_matlab_wrapper(detections, sigma_l, sigma_iou, sigma_p, sigma_len, skip_frames=False, n_skip=3):
"""
Matlab wrapper of the iou tracker for the detrac evaluation toolkit.
Args:
detections (numpy.array): numpy array of detections, usually supplied by run_tracker.m
sigma_l (float): low detection threshold.
sigma_iou (float): IOU threshold.
sigma_p (int): maximum frames a track remains pending before termination.
sigma_len (int): minimum track length in frames.
skip_frames (boolean): whether to skip some frames to speed up tracking.
n_skip (int): when skip_frames is True, the tracker uses only one out of every n_skip frames for tracking.
Returns:
float: speed in frames per second.
list: list of tracks.
"""
detections = detections.reshape((7, -1)).transpose()
detections = load_mot(detections)
start = time()
tracks = track_iou(detections, sigma_l, sigma_iou, sigma_p, sigma_len, skip_frames, n_skip)
end = time()
id_ = 1
out = []
for track in tracks:
for tracklet in track:
out += [float(tracklet['roi'][1]), float(tracklet['roi'][0]), float(tracklet['roi'][3] - tracklet['roi'][1]), float(tracklet['roi'][2] - tracklet['roi'][0]), float(tracklet['frame']), float(id_)]
id_ += 1
num_frames = len(detections)
# this part occasionally throws ZeroDivisionError when evaluated in the DETRAC toolkit without the except clause
try:
speed = num_frames / (end - start)
except:
speed = num_frames / 0.1
return speed, out
def main(args):
with open(args.seqmap) as fd:
seqs = [line.rstrip('\n') for line in fd]
for idx, seq in enumerate(seqs):
if seq == "name" or seq == "":
continue
else:
det_path = args.benchmark_dir + "/" + seq + "/det/det.txt"
out_path = args.res_dir + "/" + seq + ".txt"
detections = load_mot(det_path, with_classes=False)
start = time()
tracks = track_iou(detections, args.sigma_l, args.sigma_h,
args.sigma_iou, args.t_min)
end = time()
num_frames = len(detections)
print("finished " + seq + " at " +
str(int(num_frames / (end - start))) + " fps!")
save_to_csv(out_path, tracks)
from iou_tracker import Tracker
from util import load_mot, save_to_csv
DETS_PATH = "./MOT17/train/MOT17-04-SDP/det/det.txt"
detections = load_mot(DETS_PATH)
tracker = Tracker()
tracks = {}
for frame, detection in enumerate(detections, start=1):
track = tracker.track(detection)
tracks[frame] = track
def main(args):
time_iou = 0
time_ext = 0
time_merge = 0
time_ttl = 0
with open(args.seqmap) as fd:
seqs = [line.rstrip('\n') for line in fd]
for idx, seq in enumerate(seqs):
print(seq)
if seq == "name" or seq == "":
continue
else:
if "DPM" in seq:
sigma_l = -0.5
sigma_h = 0.5
sigma_iou = 0.5
t_min = 4
elif "FRCNN" in seq:
sigma_l = 0.0
sigma_h = 0.9
sigma_iou = 0.4
t_min = 3
elif "SDP" in seq:
sigma_l = 0.4
sigma_h = 0.5
sigma_iou = 0.3
t_min = 2
else:
sigma_l = -0.5 # default -0.5
sigma_h = 0.5 # default 0.5
sigma_iou = 0.4 # default 0.4
t_min = 3 # default 4
ttl_vtracking = 8 # maximum length of visual track (amount of framess)
sigma_iou_merge = 0.4
# uncomment line below if f4k2013 data is used
# det_path = os.path.join(args.benchmark_dir,seq+"det.txt")
# motchallenge data used
det_path = os.path.join(args.benchmark_dir, seq, "det", "det.txt")
img_path = os.path.join(args.benchmark_dir, seq, "img1")
out_path = os.path.join(args.res_dir, seq + ".txt")
detections = load_mot(det_path)
# test preprocessing detections
if args.prep_detections:
detections = detections_preprocessing(detections,
iou_filter=0.90)
start = time()
tracks_iou = track_iou(detections, sigma_l, sigma_h, sigma_iou,
t_min)
end_iou = time()
tracks_iou_ext, tracks_kcf_front, tracks_kcf_rear = track_kcf_2(
tracks_iou, img_path, ttl_vtracking)
end_ext = time()
tracks_merged = merge(tracks_iou_ext, tracks_kcf_front,
tracks_kcf_rear, sigma_iou_merge)
end = time()
num_frames = len(detections)
print("finished " + seq + " at " +
str(int(num_frames / (end - start))) + " fps!")
save_to_csv(out_path, tracks_merged)
time_iou += (end_iou - start)
time_ext += (end_ext - end_iou)
time_merge += (end - end_ext)
time_ttl += (end - start)
print('Time IOU: ' + str(time_iou))
print('Time VOT: ' + str(time_ext))
print('Time Merge:' + str(time_merge))
print('Time total: ' + str(time_ttl))
print("Reading Video {}".format(INPUT_VIDEO))
input_video = skvideo.io.vread(INPUT_VIDEO)
print("Reading Finished")
for i, frame in enumerate(input_video, start=1):
tracks = tracks_per_frame.get(i, {})
output_frame = render_frame(frame, tracks)
output_video.writeFrame(output_frame)
print("Writen Frame: {}".format(i))
output_video.close()
font = cv2.FONT_HERSHEY_DUPLEX
font_size = 0.8
if __name__ == '__main__':
#INPUT_VIDEO = "./MOT17-04-SDP.mp4"
INPUT_PATH = "MOT17/train/MOT17-04-SDP/img1"
INPUT_DETECTION = "MOT17/train/MOT17-04-SDP/det/det.txt"
OUTPUT_VIDEO = "./result.mp4"
detections = load_mot(INPUT_DETECTION)
tracker = Tracker()
tracks = {}
for frame, detection in enumerate(detections, start=1):
track = tracker.track(detection)
tracks[frame] = track
parsed_tracks = parse_tracks(tracks)
write_video(INPUT_PATH, parsed_tracks, OUTPUT_VIDEO)
