open-mmlab/mmdetection · error · RuntimeError

imagecorruptions is not installed

Error message

imagecorruptions is not installed

What it means

PhotoMetricDistortion-like Corrupt transform (mmdet) requires the optional `imagecorruptions` package to apply synthetic corruptions (gaussian noise, fog, etc.). At runtime, if the lazy import failed, the transform raises RuntimeError immediately.

Source

Thrown at mmdet/datasets/transforms/transforms.py:1530

        severity (int): The severity of corruption. Defaults to 1.
    """

    def __init__(self, corruption: str, severity: int = 1) -> None:
        self.corruption = corruption
        self.severity = severity

    def transform(self, results: dict) -> dict:
        """Call function to corrupt image.

        Args:
            results (dict): Result dict from loading pipeline.

        Returns:
            dict: Result dict with images corrupted.
        """

        if corrupt is None:
            raise RuntimeError('imagecorruptions is not installed')
        results['img'] = corrupt(
            results['img'].astype(np.uint8),
            corruption_name=self.corruption,
            severity=self.severity)
        return results

    def __repr__(self) -> str:
        repr_str = self.__class__.__name__
        repr_str += f'(corruption={self.corruption}, '
        repr_str += f'severity={self.severity})'
        return repr_str


@TRANSFORMS.register_module()
@avoid_cache_randomness
class Albu(BaseTransform):
    """Albumentation augmentation.

View on GitHub (pinned to cfd5d3a985)

Solutions

1. pip install imagecorruptions
2. Verify import: python -c "from imagecorruptions import corrupt"
3. Remove the Corrupt transform if robustness testing is not intended

Example fix

# before
pipeline=[dict(type='Corrupt', corruption='fog', severity=3)]
# after
pip install imagecorruptions  # keep the pipeline unchanged

import importlib.util
assert importlib.util.find_spec('imagecorruptions') is not None, 'install imagecorruptions'

try:
    from mmdet.datasets.transforms import Corrupt
    out = Corrupt(corruption='fog', severity=3).transform(results)
except RuntimeError as e:
    if 'imagecorruptions' in str(e):
        raise SystemExit('pip install imagecorruptions') from e

When it happens

Trigger: Using a pipeline config containing dict(type='Corrupt', corruption='gaussian_noise', severity=...) in testing pipelines (e.g. robustness benchmarks on COCO-C) without imagecorruptions installed.

Common situations: Running corruption/robustness test configs (COCO-C, VOC-C) in a fresh environment; extra deps are not installed by default `pip install mmdet`.

Related errors

• albumentations is not installed
• If you want to reduce GPU memory usage,
• If you want to reduce GPU memory usage,
• If you want to reduce GPU memory usage,
• mmlvis is deprecated, please install official lvis-api by "p

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/1ce279181303578c. Report an issue: GitHub.