def __init__(self, model, device, max_dist=0.2, max_iou_distance=0.7, max_age=70, n_init=3, nn_budget=100):
# models trained on: market1501, dukemtmcreid and msmt17
if is_model_in_factory(model):
# download the model
model_path = join('deep_sort/deep/checkpoint', model + '.pth')
if not file_exists(model_path):
gdown.download(get_model_link(model), model_path, quiet=False)
self.extractor = FeatureExtractor(
# get rid of dataset information DeepSort model name
model_name=model.rsplit('_', 1)[:-1][0],
model_path=model_path,
device=str(device)
)
else:
if is_model_type_in_model_path(model):
model_name = get_model_type(model)
self.extractor = FeatureExtractor(
model_name=model_name,
model_path=model,
device=str(device)
)
else:
print('Cannot infere model name from provided DeepSort path, should be one of the following:')
show_supported_models()
exit()
self.max_dist = max_dist
metric = NearestNeighborDistanceMetric(
"cosine", self.max_dist, nn_budget)
self.tracker = Tracker(
metric, max_iou_distance=max_iou_distance, max_age=max_age, n_init=n_init)
def main():
args = parse_args()
pwd = 'sshpass -p sihan123 '
scp = 'scp [email protected]:/home/sihan/Documents/motoolkit/reid/build/install/bin/'
os.system(pwd + scp + 'in.bin {}'.format(args.files_path))
os.system(pwd + scp + 'out.bin {}'.format(args.files_path))
extractor = FeatureExtractor(model_name='osnet_x1_0',
model_path=args.model_path,
device='cuda')
with open(osp.join(args.files_path, 'in.bin'), 'rb') as fd:
input = np.fromfile(fd, dtype=np.float32)
input = torch.from_numpy(input).view(args.batch_size, 3, 256,
128).cuda()
baseline_output = extractor(input).view(-1)
print('baseline_output:\n{}'.format(baseline_output))
with open(osp.join(args.files_path, 'out.bin'), 'rb') as fd:
test_output = np.fromfile(fd, dtype=np.float32)
test_output = torch.from_numpy(test_output).cuda()
print('test_output:\n{}'.format(test_output))
error = torch.abs(baseline_output - test_output)
print('error: min {}, max {}, mean {}'.format(error.min(), error.max(),
error.mean()))
def __init__(self, model_name: str, checkpoint_path: str,
img_size: Tuple[int, int], **kwargs):
super().__init__(**kwargs)
self.extractor = FeatureExtractor(model_name=model_name,
model_path=checkpoint_path,
device='cuda',
image_size=(img_size[1],
img_size[0]))
def init_models(video_size=(704, 576)):
yolo = YoloV5()
reid = FeatureExtractor(
model_name='osnet_x1_0',
model_path=
'pid/deep_person_reid/checkpoints/osnet_x1_0_market_256x128_amsgrad_ep150_stp60_lr0.0015_b64_fb10_softmax_labelsmooth_flip.pth',
verbose=False)
face_detector = RetinaFace(image_size=(video_size[1], video_size[0]))
# detector = MTCNN()
faceNet = FaceNet()
return yolo, reid, face_detector, faceNet
def __init__(self, proxy_map, startup_check=False):
super(SpecificWorker, self).__init__(proxy_map)
self.Period = 2000
if startup_check:
self.startup_check()
else:
self.timer.timeout.connect(self.compute)
self.timer.start(self.Period)
# Load center track, our tracking module
opt = opts().init()
self.detector = Detector(opt)
# Reid reture extractor
self.reid_extractor = FeatureExtractor(
model_name='shufflenet',
model_path='/home/shubh/Downloads/shufflenet-bee1b265.pth.tar',
device='cuda')
def __init__(self,
model_type,
device,
max_dist=0.2,
max_iou_distance=0.7,
max_age=70,
n_init=3,
nn_budget=100):
self.extractor = FeatureExtractor(model_name=model_type,
device=str(device))
max_cosine_distance = max_dist
metric = NearestNeighborDistanceMetric("cosine", max_cosine_distance,
nn_budget)
self.tracker = Tracker(metric,
max_iou_distance=max_iou_distance,
max_age=max_age,
n_init=n_init)
def make_model():
yolo = yolo_model()
reid = FeatureExtractor(
model_name='osnet_x1_0',
model_path=
'pid/deep_person_reid/checkpoints/osnet_x1_0_market_256x128_amsgrad_ep150_stp60_lr0.0015_b64_fb10_softmax_labelsmooth_flip.pth',
verbose=False)
pnet, rnet, onet = create_mtcnn_net(
p_model_path="fid/mtcnn/mtcnn_checkpoints/pnet_epoch.pt",
r_model_path="fid/mtcnn/mtcnn_checkpoints/rnet_epoch.pt",
o_model_path="fid/mtcnn/mtcnn_checkpoints/onet_epoch.pt",
use_cuda=True)
mtcnn_detector = MtcnnDetector(pnet=pnet,
rnet=rnet,
onet=onet,
min_face_size=24)
mobileFace = mobile_face_model(
"fid/InsightFace_Pytorch/facenet_checkpoints/model_ir_se50.pth")
return yolo, reid, mtcnn_detector, mobileFace
# -*- coding: utf-8 -*- # !@time: 2020/6/9 上午5:42 # !@author: superMC @email: [email protected] # !@fileName: pid_demo.py import cv2 from torchreid.utils import FeatureExtractor from pid.yolov4.yolov4 import YoloV4 as Yolo #from pid.yolov5.yolov5 import YoloV5 as Yolo if __name__ == '__main__': import time yolo = Yolo() reid = FeatureExtractor( model_name='osnet_x1_0', model_path= 'pid/deep_person_reid/checkpoints/osnet_x1_0_market_256x128_amsgrad_ep150_stp60_lr0.0015_b64_fb10_softmax_labelsmooth_flip.pth', verbose=False) image = cv2.imread("data/office1.jpg") # person_images, _ = detect_person(model, image) person_images, person_boxes = yolo(image) for box in person_boxes: image = cv2.rectangle(image, (box[0], box[1]), (box[2], box[3]), (255, 0, 0), 2) cv2.imshow('demo', image) cv2.waitKey(0)
def _set_extractor(self):
self.extractor = FeatureExtractor(model_name=self.model_name,
model_path=self.model_path,
device=self.device)
# Arguments Parser
parser = argparse.ArgumentParser()
parser.add_argument("-v",
"--video",
type=str,
help="Surviellance video file path")
parser.add_argument("-s",
"--save",
type=bool,
default=True,
help="Save cropped Images of Pedestrain")
args = parser.parse_args()
# OSNet Feature Extractor
extractor = FeatureExtractor(
model_name='osnet_x1_0',
model_path='model_weights/osnet_weights.tar-250',
device='cuda')
model_path = 'model_weights/ssd_weights.pb'
odapi = PersonDetector(path_to_ckpt=model_path)
threshold = 0.7
cap = cv2.VideoCapture(args.video)
frame_counter = 0
# Make a directory for saving cropped images of pedestrian
gallery_path = "Gallery"
os.makedirs(gallery_path, exist_ok=True)
os.chdir(gallery_path)
Features = []
def main():
args = parse_args()
if osp.isfile(args.config):
config.merge_from_file(args.config)
config.freeze()
calibs = [parse_calibration_data(f) for f in args.calibration]
print('calibration:\n{}'.format(calibs))
# Create input and output queue for each tracker.
ncamera = len(args.inputs)
images = [] # image queues
tracklets = [] # tracklet queues
for i in range(ncamera):
images.append(Queue(maxsize=0))
tracklets.append(Queue(maxsize=0))
trajectories = [] # global trajectory list
# Feature extractor.
extractor = FeatureExtractor(model_name='osnet_x1_0',
model_path=args.reid,
device='cuda')
# Create working threads.
tid = 0
threads = []
exit = Value('i', 0) # shared thread exit switch
for i in range(ncamera):
# Datastore thread.
tid += 1
threads.append(Datastore(tid, args.inputs[i], images[i]))
# MTSCT thread.
tid += 1
model = build_tracker(config.MODEL)
model.load_state_dict(torch.load(args.tracker, map_location='cpu'))
locator = ImageToWorldTsai(calibs[i])
threads.append(
MTSCT(tid, images[i], tracklets[i], exit, model, locator))
# MTMCT thread.
tid += 1
threads.append(MTMCT(tid, tracklets, trajectories, exit, extractor))
# Start all threads.
for thread in threads:
thread.start()
# Waiting for Datastore finish.
ndead = 0
while ndead != ncamera:
ndead = sum([int(not t.is_alive()) for t in threads[:-1][0::2]])
time.sleep(1)
print('Datastore done.')
# Waiting for MTSCT finish.
nempty = 0
while nempty != ncamera:
nempty = sum([int(q.empty()) for q in images])
time.sleep(1)
print('MTSCT done.')
# Waiting for MTMCT finish.
nempty = 0
while nempty != ncamera:
nempty = sum([int(q.empty()) for q in tracklets])
time.sleep(1)
print('MTMCT done.')
exit.value = 1
for thread in threads:
thread.join()
print('All works done.')
from torchreid.utils import FeatureExtractor
import time
extractor = FeatureExtractor(
model_name='osnet_x0_25',
model_path='torchreid/models/osnet_x0_25_imagenet.pth',
device='cpu')
image_list = ['torchreid/data/img.png']
total_time = 0
for i in range(100):
time1 = time.time()
features = extractor(image_list)
time2 = time.time()
total_time += (time2 - time1)
#print(features.shape) # output (N, 512)
avg_time = total_time / len(range(100))
print(avg_time, "Avg time per loop")
print(1 / avg_time, "Avg FPS")