opendatalab/MinerU · error · ValueError

Invalid scale {scale}, must be positive.

Error message

Invalid scale {scale}, must be positive.

What it means

Raised by rescale_size() (mmcv-derived helper inside mineru's UNet table recognizer) when the scale argument is a numeric value that is zero or negative. The function multiplies both image edges by this factor, so a non-positive factor is mathematically meaningless and would produce a zero-size image. It exists to fail fast before cv2/NumPy produce a cryptic downstream error.

Source

Thrown at mineru/model/table/rec/unet_table/utils.py:317

def rescale_size(old_size, scale, return_scale=False):
    """Calculate the new size to be rescaled to.

    Args:
        old_size (tuple[int]): The old size (w, h) of image.
        scale (float | tuple[int]): The scaling factor or maximum size.
            If it is a float number, then the image will be rescaled by this
            factor, else if it is a tuple of 2 integers, then the image will
            be rescaled as large as possible within the scale.
        return_scale (bool): Whether to return the scaling factor besides the
            rescaled image size.

    Returns:
        tuple[int]: The new rescaled image size.
    """
    w, h = old_size
    if isinstance(scale, (float, int)):
        if scale <= 0:
            raise ValueError(f"Invalid scale {scale}, must be positive.")
        scale_factor = scale
    elif isinstance(scale, tuple):
        max_long_edge = max(scale)
        max_short_edge = min(scale)
        scale_factor = min(max_long_edge / max(h, w), max_short_edge / min(h, w))
    else:
        raise TypeError(
            f"Scale must be a number or tuple of int, but got {type(scale)}"
        )

    new_size = _scale_size((w, h), scale_factor)

    if return_scale:
        return new_size, scale_factor
    else:
        return new_size

View on GitHub (pinned to 4fe4bde114)

Solutions

  1. Check the scale value passed to the table preprocessing config; it must be a positive float/int or a tuple like (736, 1280).
  2. If scale is computed dynamically, guard the denominator: scale = target / max(max(h, w), 1).
  3. Print/inspect the input image size right before the call; a (0, 0) size usually means the image failed to load earlier.
  4. If you intended a max-size constraint, pass the tuple form (short_edge, long_edge) instead of a single factor.

Example fix

// before
new_size = rescale_size(old_size, 0)  # scale typo / computed as 0

// after
new_size = rescale_size(old_size, (736, 1280))  # tuple form, or a positive factor
if isinstance(scale, (int, float)) and scale <= 0:
    raise ValueError(f"bad scale config: {scale}")
Defensive patterns

Strategy: validation

Validate before calling

def safe_scale(scale):
    if isinstance(scale, (int, float)) and scale <= 0:
        raise ValueError(f"scale must be positive, got {scale}")
    return scale

new_size = rescale_size(size, safe_scale(cfg_scale))

Type guard

def is_valid_scale(s) -> bool:
    return (isinstance(s, (int, float)) and s > 0) or (
        isinstance(s, tuple) and len(s) == 2 and all(isinstance(v, int) for v in s)
    )

Try / catch

try:
    new_size = rescale_size(size, scale)
except ValueError as e:
    if "must be positive" in str(e):
        scale = 1.0
        new_size = rescale_size(size, scale)
    else:
        raise

Prevention

When it happens

Trigger: Calling rescale_size(old_size, scale) (or the table-rec pipeline that calls it, e.g. table structure preprocessing) with scale=0, a negative float, or a value computed from division that underflowed to 0, e.g. scale = target_px / max(h, w) when max(h, w) is huge or target_px is 0.

Common situations: Custom table-detection configs that pass a user-defined scale factor of 0 by mistake; dynamically computed ratios where the denominator is an image whose dimensions were read as 0 (corrupt image or failed load); copy-pasting a config where scale was meant to be a (short, long) tuple but only one element was supplied.

Related errors

AI-assisted analysis of opendatalab/MinerU@4fe4bde114 (2026-08-14). Data as JSON: /api/errors/a45cfd4646dde288. Report an issue: GitHub.