open-mmlab/mmdetection · error · RuntimeError

skimage is not installed, please install it

Error message

skimage is not installed,                 please install it by: pip install scikit-image

What it means

BatchSyncRandomResizeFlipColorJitter or similar data preprocessor (BatchAugment) requires scikit-image for its color-similarity augmentation (pairwise color term used by mask-aware photometric aug). If the optional import failed, __init__ raises RuntimeError with install instructions.

Source

Thrown at mmdet/models/data_preprocessors/data_preprocessor.py:699

    """

    def __init__(self,
                 *arg,
                 mask_stride: int = 4,
                 pairwise_size: int = 3,
                 pairwise_dilation: int = 2,
                 pairwise_color_thresh: float = 0.3,
                 bottom_pixels_removed: int = 10,
                 **kwargs) -> None:
        super().__init__(*arg, **kwargs)
        self.mask_stride = mask_stride
        self.pairwise_size = pairwise_size
        self.pairwise_dilation = pairwise_dilation
        self.pairwise_color_thresh = pairwise_color_thresh
        self.bottom_pixels_removed = bottom_pixels_removed

        if skimage is None:
            raise RuntimeError('skimage is not installed,\
                 please install it by: pip install scikit-image')

    def get_images_color_similarity(self, inputs: Tensor,
                                    image_masks: Tensor) -> Tensor:
        """Compute the image color similarity in LAB color space."""
        assert inputs.dim() == 4
        assert inputs.size(0) == 1

        unfolded_images = unfold_wo_center(
            inputs,
            kernel_size=self.pairwise_size,
            dilation=self.pairwise_dilation)
        diff = inputs[:, :, None] - unfolded_images
        similarity = torch.exp(-torch.norm(diff, dim=1) * 0.5)

        unfolded_weights = unfold_wo_center(
            image_masks[None, None],
            kernel_size=self.pairwise_size,

View on GitHub (pinned to cfd5d3a985)

Solutions

1. pip install scikit-image
2. Or remove/disable the color-jitter preprocessor from the config
3. Reinstall mmdet with extras: mim install 'mmdet[extra]' style dependency set

Example fix

// shell
pip install scikit-image

try:\n    import skimage  # noqa\nexcept ImportError:\n    raise SystemExit('pip install scikit-image before using color-jitter preprocessor')

When it happens

Trigger: Constructing the data preprocessor with pairwise_color_thresh / photometric distortion options without scikit-image installed (pip install did not include the extra).

Common situations: Running RTMDet/SOLO training configs that use BatchSyncRandomResizeFlipColorJitter in a minimal env; fresh install without extras.

Related errors

• Please run "pip install openmim" and run "mim install mmpret
• panopticapi is not installed, please install it by: pip inst
• The annotation file of Open Images Challenge should be a txt
• Invalid text mode "{self.text_mode}".
• No sample in split "{self.split}".

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/19691e6d040b9243. Report an issue: GitHub.