def test_tracker_by_name(tracker, similarity, match, count):
t = Tracker.make_tracker_by_name("flow",
"instance",
"greedy",
clean_instance_count=2)
t.track([])
t.final_pass([])
def make_cli_parser():
import argparse
from sleap.util import frame_list
parser = argparse.ArgumentParser()
# Add args for entire pipeline
parser.add_argument(
"video_path", type=str, nargs="?", default="", help="Path to video file"
)
parser.add_argument(
"-m",
"--model",
dest="models",
action="append",
help="Path to trained model directory (with training_config.json). "
"Multiple models can be specified, each preceded by --model.",
)
parser.add_argument(
"--frames",
type=frame_list,
default="",
help="List of frames to predict. Either comma separated list (e.g. 1,2,3) or "
"a range separated by hyphen (e.g. 1-3, for 1,2,3). (default is entire video)",
)
parser.add_argument(
"--only-labeled-frames",
action="store_true",
default=False,
help="Only run inference on labeled frames (when running on labels dataset file).",
)
parser.add_argument(
"--only-suggested-frames",
action="store_true",
default=False,
help="Only run inference on suggested frames (when running on labels dataset file).",
)
parser.add_argument(
"-o",
"--output",
type=str,
default=None,
help="The output filename to use for the predicted data.",
)
parser.add_argument(
"--labels",
type=str,
default=None,
help="Path to labels dataset file (for inference on multiple videos or for re-tracking pre-existing predictions).",
)
# TODO: better video parameters
parser.add_argument(
"--video.dataset", type=str, default="", help="The dataset for HDF5 videos."
)
parser.add_argument(
"--video.input_format",
type=str,
default="",
help="The input_format for HDF5 videos.",
)
device_group = parser.add_mutually_exclusive_group(required=False)
device_group.add_argument(
"--cpu",
action="store_true",
help="Run inference only on CPU. If not specified, will use available GPU.",
)
device_group.add_argument(
"--first-gpu",
action="store_true",
help="Run inference on the first GPU, if available.",
)
device_group.add_argument(
"--last-gpu",
action="store_true",
help="Run inference on the last GPU, if available.",
)
device_group.add_argument(
"--gpu", type=int, default=0, help="Run inference on the i-th GPU specified."
)
# Add args for each predictor class
for predictor_name, predictor_class in CLI_PREDICTORS.items():
if "peak_threshold" in attr.fields_dict(predictor_class):
# get the default value to show in help string, although we'll
# use None as default so that unspecified vals won't be passed to
# builder.
default_val = attr.fields_dict(predictor_class)["peak_threshold"].default
parser.add_argument(
f"--{predictor_name}.peak_threshold",
type=float,
default=None,
help=f"Threshold to use when finding peaks in {predictor_class.__name__} (default: {default_val}).",
)
if "batch_size" in attr.fields_dict(predictor_class):
default_val = attr.fields_dict(predictor_class)["batch_size"].default
parser.add_argument(
f"--{predictor_name}.batch_size",
type=int,
default=None,
help=f"Batch size to use for model inference in {predictor_class.__name__} (default: {default_val}).",
)
# Add args for tracking
Tracker.add_cli_parser_args(parser, arg_scope="tracking")
parser.add_argument(
"--test-pipeline",
default=False,
action="store_true",
help="Test pipeline construction without running anything.",
)
return parser
def make_tracker_from_cli(policy_args):
if "tracking" in policy_args:
tracker = Tracker.make_tracker_by_name(**policy_args["tracking"])
return tracker
return None