open-mmlab/mmdetection · error · ValueError

Invalid crop_type {crop_type}.

Error message

Invalid crop_type {crop_type}.

What it means

RandomCrop's __init__ validates crop_type against the four supported modes: 'relative_range', 'relative', 'absolute', 'absolute_range'. Any other string (typos like 'abs', 'relative-range', 'Absolute', or empty) raises ValueError immediately at pipeline construction.

Source

Thrown at mmdet/datasets/transforms/transforms.py:858

            original image.
        - The keys for bboxes, labels and masks must be aligned. That is,
          ``gt_bboxes`` corresponds to ``gt_labels`` and ``gt_masks``, and
          ``gt_bboxes_ignore`` corresponds to ``gt_labels_ignore`` and
          ``gt_masks_ignore``.
        - If the crop does not contain any gt-bbox region and
          ``allow_negative_crop`` is set to False, skip this image.
    """

    def __init__(self,
                 crop_size: tuple,
                 crop_type: str = 'absolute',
                 allow_negative_crop: bool = False,
                 recompute_bbox: bool = False,
                 bbox_clip_border: bool = True) -> None:
        if crop_type not in [
                'relative_range', 'relative', 'absolute', 'absolute_range'
        ]:
            raise ValueError(f'Invalid crop_type {crop_type}.')
        if crop_type in ['absolute', 'absolute_range']:
            assert crop_size[0] > 0 and crop_size[1] > 0
            assert isinstance(crop_size[0], int) and isinstance(
                crop_size[1], int)
            if crop_type == 'absolute_range':
                assert crop_size[0] <= crop_size[1]
        else:
            assert 0 < crop_size[0] <= 1 and 0 < crop_size[1] <= 1
        self.crop_size = crop_size
        self.crop_type = crop_type
        self.allow_negative_crop = allow_negative_crop
        self.bbox_clip_border = bbox_clip_border
        self.recompute_bbox = recompute_bbox

    def _crop_data(self, results: dict, crop_size: Tuple[int, int],
                   allow_negative_crop: bool) -> Union[dict, None]:
        """Function to randomly crop images, bounding boxes, masks, semantic
        segmentation maps.

View on GitHub (pinned to cfd5d3a985)

Solutions

1. Set crop_type to one of 'relative_range' (default), 'relative', 'absolute', or 'absolute_range'
2. Remember semantics: 'relative*' treats crop_size as fractions of image size, 'absolute*' as pixels, 'absolute_range' picks a random size up to crop_size

Example fix

# before
dict(type='RandomCrop', crop_size=(0.5, 0.5), crop_type='relative-range')
# after
dict(type='RandomCrop', crop_size=(0.5, 0.5), crop_type='relative_range')

VALID = {'relative_range', 'relative', 'absolute', 'absolute_range'}
assert cfg['crop_type'] in VALID, f"crop_type must be one of {VALID}"

def is_valid_crop_type(ct: str) -> bool:
    return ct in {'relative_range', 'relative', 'absolute', 'absolute_range'}

When it happens

Trigger: Building dict(type='RandomCrop', crop_size=(w, h), crop_type='...') with a crop_type not exactly one of the four allowed lowercase strings; the check is an exact membership test so case and spelling must match.

Common situations: Hand-written configs with typo'd or translated crop_type values; porting crop configs from other libraries (e.g. albumentations 'px'/'percent' vocabulary) into mmdet; copy-paste from outdated docs.

Understand the failure class

Background: Invalid enum value errors: "Unknown type", "Invalid scope", "must be one of" — when a string is not on the library's allowed list — this error's family across 23 libraries.

Related errors

• Invalid scale {scale}, must be positive.
• Please run "pip install instaboostfast" to install instaboos
• Please run "pip install instaboostfast" to install instaboos
• Scale must be a number or tuple of int, but got {type(scale)
• basesize_ratio_range[0] should be either 0.15or 0.2 when inp

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/376b2a28317f0844. Report an issue: GitHub.