keras-team/keras · error · ValueError

Invalid axis' indices: {actual_indices - valid_indices}

Error message

Invalid axis' indices: {actual_indices - valid_indices}

What it means

apply_affine_transform only accepts axis indices in {0,1,2} because it operates on 3D (multi-channel 2D) arrays. After checking the axes are distinct it verifies the index set equals {0,1,2} and raises this error listing the offending indices. Negative indices such as -1 are explicitly unsupported.

Source

Thrown at keras/src/legacy/preprocessing/image.py:1803

    order=1,
):
    """Applies an affine transformation specified by the parameters given.

    DEPRECATED.
    """
    # Input sanity checks:
    # 1. x must 2D image with one or more channels (i.e., a 3D tensor)
    # 2. channels must be either first or last dimension
    if np.unique([row_axis, col_axis, channel_axis]).size != 3:
        raise ValueError(
            "'row_axis', 'col_axis', and 'channel_axis' must be distinct"
        )

    # shall we support negative indices?
    valid_indices = set([0, 1, 2])
    actual_indices = set([row_axis, col_axis, channel_axis])
    if actual_indices != valid_indices:
        raise ValueError(
            f"Invalid axis' indices: {actual_indices - valid_indices}"
        )

    if x.ndim != 3:
        raise ValueError("Input arrays must be multi-channel 2D images.")
    if channel_axis not in [0, 2]:
        raise ValueError(
            "Channels are allowed and the first and last dimensions."
        )

    transform_matrix = None
    if theta != 0:
        theta = np.deg2rad(theta)
        rotation_matrix = np.array(
            [
                [np.cos(theta), -np.sin(theta), 0],
                [np.sin(theta), np.cos(theta), 0],
                [0, 0, 1],

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Replace negative indices: use 2 instead of -1 for channels-last
2. Ensure input x is a single 3D image (H, W, C); loop over the batch dimension yourself
3. Confirm all three axes form a permutation of 0, 1, 2

Example fix

# before
apply_affine_transform(img, tx=2, row_axis=0, col_axis=1, channel_axis=-1)
# after
apply_affine_transform(img, tx=2, row_axis=0, col_axis=1, channel_axis=2)

if any(a not in (0, 1, 2) for a in (row_axis, col_axis, channel_axis)):
    raise ValueError('axes must be in 0..2; negative indices unsupported')

def valid_axes(r, c, ch):
    return all(isinstance(a, int) and 0 <= a <= 2 for a in (r, c, ch))

When it happens

Trigger: Passing row_axis=-1 or channel_axis=3 to apply_affine_transform or the random_* wrappers; passing axes meant for a 4D batch tensor (axis=3) while feeding a single image.

Common situations: Porting numpy/scipy-style code that uses negative axis indices; feeding a batched NHWC tensor and giving axis=3 for channels.

Understand the failure class

Background: "Must be a positive integer", "Invalid value", "Unsupported": the invalid-argument-value error family, when a library rejects the value you pass — this error's family across 35 libraries.

Related errors

• 'row_axis', 'col_axis', and 'channel_axis' must be distinct
• The `factor` argument should be a number (or a list of two n
• Received: {factor_name}={factor}
• Received: {factor_name}={factor}
• Received: value_range={value_range}

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/f5bbabbdbed5da34. Report an issue: GitHub.