open-mmlab/mmdetection · error · RuntimeError

sscikit-learn is not installed, please insta

Error message

sscikit-learn is not installed,                 please install it by: pip install scikit-learn

What it means

The TrackletInterpolator (used for video instance/seg tracking interpolation between sparse keyframes) requires Gaussian-smoothed interpolation backed by scikit-learn. At import time mmdet tries to import sklearn; if it is missing, HAS_SKIKIT_LEARN is False and __init__ raises RuntimeError.

Source

Thrown at mmdet/models/task_modules/tracking/interpolation.py:34

    """Interpolate tracks to make tracks more complete.

    Args:
        min_num_frames (int, optional): The minimum length of a track that will
            be interpolated. Defaults to 5.
        max_num_frames (int, optional): The maximum disconnected length in
            a track. Defaults to 20.
        use_gsi (bool, optional): Whether to use the GSI (Gaussian-smoothed
            interpolation) method. Defaults to False.
        smooth_tau (int, optional): smoothing parameter in GSI. Defaults to 10.
    """

    def __init__(self,
                 min_num_frames: int = 5,
                 max_num_frames: int = 20,
                 use_gsi: bool = False,
                 smooth_tau: int = 10):
        if not HAS_SKIKIT_LEARN:
            raise RuntimeError('sscikit-learn is not installed,\
                 please install it by: pip install scikit-learn')
        self.min_num_frames = min_num_frames
        self.max_num_frames = max_num_frames
        self.use_gsi = use_gsi
        self.smooth_tau = smooth_tau

    def _interpolate_track(self,
                           track: np.ndarray,
                           track_id: int,
                           max_num_frames: int = 20) -> np.ndarray:
        """Interpolate a track linearly to make the track more complete.

        This function is proposed in
        "ByteTrack: Multi-Object Tracking by Associating Every Detection Box."
        `ByteTrack<https://arxiv.org/abs/2110.06864>`_.

        Args:
            track (ndarray): With shape (N, 7). Each row denotes

View on GitHub (pinned to cfd5d3a985)

Solutions

  1. pip install scikit-learn
  2. Or verify with 'python -c "import sklearn"' that it's importable in the same env as mmdet
  3. If using conda: conda install -c conda-forge scikit-learn

Example fix

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

Strategy: validation

Validate before calling

try:
    import sklearn  # noqa
    ok = True
except ImportError:
    ok = False
assert ok, 'pip install scikit-learn before using tracking interpolation'

Try / catch

try:
    interp = TrackletInterpolator(...)
except RuntimeError as e:
    if 'scikit-learn' in str(e):
        raise SystemExit('Missing dependency: run pip install scikit-learn')
    raise

Prevention

When it happens

Trigger: Instantiating TrackletInterpolator with use_gsi=True workflows or simply constructing it in any environment where 'import sklearn' failed (extra deps not installed).

Common situations: Installing mmdet with minimal deps (pip install mmdet without [extra]) and then running video tracking / VIS configs (e.g. Mask2Former video) that build this module.

Related errors

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/cb6f2c571969778e. Report an issue: GitHub.