confidence=confidence, x1=int(detection[3] * img_width), y1=int(detection[4] * img_ht),
x2=int(detection[5] * img_width), y2=int(detection[6] * img_ht))
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--grpc", default=False, help="If true, this analytic will set up a gRPC service instead of a REST service.", action="store_true")
parser.add_argument("--grpc_port", default=50051, help="Port the analytic will run on.")
parser.add_argument("--model", default="models/ssd_mobilenet_v2_coco_2018_03_29/frozen_inference_graph.pb", help="Model file to load")
parser.add_argument("--model_config", default="models/ssd_mobilenet_v2_coco_2018_03_29/ssd_mobilenet_v2_coco_2018_03_29.pbtxt", help="Model config file location")
parser.add_argument("--classes", default="coco.json", help="JOSN file mapping output vector to class names")
parser.add_argument("--verbose", "-v", default=False, help="Display additional output.", action="store_true")
parser.add_argument("--confidence_threshold", default=0.5, help="Confidence threshold for detection. Any object with a confidence socre less than this will not be considered a detection. Default 0.5")
args = parser.parse_args()
confThreshold = args.confidence_threshold
with open(args.classes, 'r') as f:
classes = json.load(f)
net = cv2.dnn.readNet(args.model, args.model_config)
if args.grpc:
svc = grpcservice.AnalyticServiceGRPC()
svc.register_name("opencv_object_detector")
svc.RegisterProcessVideoFrame(detect)
sys.exit(svc.Run(analytic_port=args.grpc_port))
else:
svc = analyticservice.AnalyticService(__name__, verbose=args.verbose)
svc.register_name("opencv_object_detector")
svc.RegisterProcessVideoFrame(detect)
sys.exit(svc.Run())
if __name__ == "__main__":
# Optional statement to configure preferred GPU. Available only in GPU version.
import argparse
import pydarknet
parser = argparse.ArgumentParser()
parser.add_argument("--host",
default="localhost",
help="Host of the proxy")
parser.add_argument("--port",
default=50051,
help="Port the proxy will run on.")
args = parser.parse_args()
# pydar knet.set_cuda_device(0)
dknet_config = {
"cfg_path": CFG,
"weights_path": WEIGHTS,
"data_path": DATA
}
logger.info(dknet_config)
net = Detector(bytes(dknet_config["cfg_path"], encoding="utf-8"),
bytes(dknet_config["weights_path"], encoding="utf-8"), 0,
bytes(dknet_config["data_path"], encoding="utf-8"))
svc = analyticservice.AnalyticService()
svc.register_name("Yolo v3")
svc.RegisterProcessVideoFrame(process_frame)
sys.exit(svc.Run(analytic_port=int(args.port)))
handler.get_analytic_metadata))
logger.info("Adding frame to response")
handler.add_frame_info(include_frame=True)
logger.info("Adding tags 'test=True' and 'LuckyNumber=7'")
handler.add_tags(test=True, LuckyNumber=7)
logger.info("Finished tests for frame")
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"--config_port",
default=3000,
help="Port the analaytic configuration endpoint runs on.")
logger = logging.getLogger(__name__)
logger.setLevel(logging.DEBUG)
formatter = logging.Formatter(
'%(asctime)s - %(name)s - %(levelname)s - %(message)s')
ch = logging.StreamHandler()
ch.setLevel(logging.DEBUG)
ch.setFormatter(formatter)
logger.addHandler(ch)
args = parser.parse_args()
svc = analyticservice.AnalyticService(__name__)
svc.register_name("test_frame_analytic")
svc.RegisterProcessFrameBatch(detect, batch_size=16)
sys.exit(svc.Run())
outputs = net.forward()
class_pred = np.argmax(outputs)
label = labels[class_pred]
handler.add_tags(activity=label)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"--config_port",
default=3000,
help="Port the analaytic configuration endpoint runs on.")
parser.add_argument("--model",
"-m",
default="resnet-34-kinetics.onnx",
help="Path to the model file to load")
parser.add_argument("--labels",
"-l",
default="action_recognition_kinetics.txt",
help="Path to class labels")
args = parser.parse_args()
net = load_net(args.model)
labels = get_labels(args.labels)
svc = analyticservice.AnalyticService(__name__, verbose=True)
svc.register_name("test_frame_analytic")
svc.RegisterProcessFrameBatch(detect, batch_size=16)
sys.exit(svc.Run())