Skip to content

Tracker Factory

Use the tracker factory helpers when you need to construct tracker backends programmatically from their registered names.

Creates and returns an instance of the specified tracker type.

Parameters: - tracker_type: The type of the tracker (e.g., 'strongsort', 'ocsort'). - tracker_config: Path to the tracker configuration file. - reid_weights: Weights for ReID (re-identification). - device: Device to run the tracker on (e.g., 'cpu', 'cuda'). - half: Boolean indicating whether to use half-precision. - per_class: Boolean for class-specific tracking (optional). - evolve_param_dict: A dictionary of parameters for evolving the tracker.

Returns: - An instance of the selected tracker.

Raises: - ValueError: If tracker_type is not recognized.

Source code in boxmot/trackers/tracker_zoo.py 
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
def create_tracker(
    tracker_type,
    tracker_config=None,
    reid_weights=None,
    device=None,
    half=None,
    per_class=None,
    evolve_param_dict=None,
):
    """
    Creates and returns an instance of the specified tracker type.

    Parameters:
    - tracker_type: The type of the tracker (e.g., 'strongsort', 'ocsort').
    - tracker_config: Path to the tracker configuration file.
    - reid_weights: Weights for ReID (re-identification).
    - device: Device to run the tracker on (e.g., 'cpu', 'cuda').
    - half: Boolean indicating whether to use half-precision.
    - per_class: Boolean for class-specific tracking (optional).
    - evolve_param_dict: A dictionary of parameters for evolving the tracker.

    Returns:
    - An instance of the selected tracker.

    Raises:
    - ValueError: If `tracker_type` is not recognized.
    """

    if tracker_type not in TRACKER_MAPPING:
        available = ", ".join(TRACKER_MAPPING.keys())
        raise ValueError(f"Unknown tracker type: '{tracker_type}'. Available trackers are: {available}")

    # Load configuration from file or use provided dictionary
    if evolve_param_dict is None:
        if tracker_config is None: 
            # Load default tracker config 
            tracker_config = get_tracker_config(tracker_type)
        with open(tracker_config, "r") as f:
            yaml_config = yaml.safe_load(f)
            tracker_args = {
                param: details["default"] for param, details in yaml_config.items()
            }
    else:
        tracker_args = evolve_param_dict.copy()

    # Prepare arguments
    tracker_args["per_class"] = per_class

    if tracker_type in REID_TRACKERS:
        tracker_args.update({
            "reid_weights": reid_weights,
            "device": device,
            "half": half,
        })

    # Tracker-specific adjustments
    if tracker_type == "strongsort":
        tracker_args.pop("per_class", None)

    # Dynamically import and instantiate the correct tracker class
    module_path, class_name = TRACKER_MAPPING[tracker_type].rsplit(".", 1)
    module = importlib.import_module(module_path)
    tracker_class = getattr(module, class_name)

    # Return the instantiated tracker class with arguments and warmed-up models
    tracker = tracker_class(**tracker_args)
    if hasattr(tracker, "model"):
        tracker.model.warmup()
    return tracker

Returns the path to the tracker configuration file.

Source code in boxmot/trackers/tracker_zoo.py 
22
23
24
def get_tracker_config(tracker_type):
    """Returns the path to the tracker configuration file."""
    return TRACKER_CONFIGS / f"{tracker_type}.yaml"