open-mmlab/mmdetection · error · RuntimeError

motmetrics is not installed, please install

Error message

motmetrics is not installed,                 please install it by: pip install motmetrics

What it means

StrongSORT tracker construction checks that motmetrics imported successfully; if not, it raises RuntimeError. StrongSORT additionally needs a ReID model and KalmanFilter, but this specific error is purely the missing motmetrics package.

Source

Thrown at mmdet/models/trackers/strongsort_tracker.py:79

            Defaults to 0.7.
        num_tentatives (int, optional): Number of continuous frames to confirm
            a track. Defaults to 2.
    """

    def __init__(self,
                 motion: Optional[dict] = None,
                 obj_score_thr: float = 0.6,
                 reid: dict = dict(
                     num_samples=None,
                     img_scale=(256, 128),
                     img_norm_cfg=None,
                     match_score_thr=0.3,
                     motion_weight=0.02),
                 match_iou_thr: float = 0.7,
                 num_tentatives: int = 2,
                 **kwargs):
        if motmetrics is None:
            raise RuntimeError('motmetrics is not installed,\
                 please install it by: pip install motmetrics')
        super().__init__(motion, obj_score_thr, reid, match_iou_thr,
                         num_tentatives, **kwargs)

    def update_track(self, id: int, obj: Tuple[Tensor]) -> None:
        """Update a track."""
        for k, v in zip(self.memo_items, obj):
            v = v[None]
            if self.momentums is not None and k in self.momentums:
                m = self.momentums[k]
                self.tracks[id][k] = (1 - m) * self.tracks[id][k] + m * v
            else:
                self.tracks[id][k].append(v)

        if self.tracks[id].tentative:
            if len(self.tracks[id]['bboxes']) >= self.num_tentatives:
                self.tracks[id].tentative = False
        bbox = bbox_xyxy_to_cxcyah(self.tracks[id].bboxes[-1])  # size = (1, 4)

View on GitHub (pinned to cfd5d3a985)

Solutions

  1. pip install motmetrics
  2. Ensure the same interpreter is used (pip from the right conda env)
  3. Install the other tracking extras (lap, scipy, scikit-learn) in one go

Example fix

# before: RuntimeError: motmetrics is not installed
# after
pip install motmetrics lap scipy scikit-learn
Defensive patterns

Strategy: validation

Validate before calling

import importlib.util
assert importlib.util.find_spec('motmetrics'), 'pip install motmetrics'

Try / catch

try:
    tracker = StrongSORTTracker(...)
except RuntimeError as e:
    if 'motmetrics' in str(e):
        raise SystemExit('Missing dependency: pip install motmetrics')
    raise

Prevention

When it happens

Trigger: tracker=dict(type='StrongSORTTracker', reid=...) built in an environment lacking py-motmetrics.

Common situations: Running StrongSORT configs (strongsort_yolox_x_8xb4-80e_mot17) on minimal mmdet installs or misconfigured conda envs.

Related errors

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/7bd6d8dcd01fd154. Report an issue: GitHub.