keras-team/keras · error · ValueError

Expected data_format to be one of `channels_first` or `chann

Error message

Expected data_format to be one of `channels_first` or `channels_last`. Received: data_format={data_format}

What it means

Runtime validation in _crop_images: only rank-3 (single image) or rank-4 (batched) image tensors are accepted; anything else raises before cropping executes.

Source

Thrown at keras/src/applications/imagenet_utils.py:98

PREPROCESS_INPUT_RET_DOC_CAFFE = """
      The images are converted from RGB to BGR, then each color channel is
      zero-centered with respect to the ImageNet dataset, without scaling."""


@keras_export("keras.applications.imagenet_utils.preprocess_input")
def preprocess_input(x, data_format=None, mode="caffe"):
    """Preprocesses a tensor or Numpy array encoding a batch of images."""
    if mode not in {"caffe", "tf", "torch"}:
        raise ValueError(
            "Expected mode to be one of `caffe`, `tf` or `torch`. "
            f"Received: mode={mode}"
        )

    if data_format is None:
        data_format = backend.image_data_format()
    elif data_format not in {"channels_first", "channels_last"}:
        raise ValueError(
            "Expected data_format to be one of `channels_first` or "
            f"`channels_last`. Received: data_format={data_format}"
        )

    if isinstance(x, np.ndarray):
        return _preprocess_numpy_input(x, data_format=data_format, mode=mode)
    else:
        return _preprocess_tensor_input(x, data_format=data_format, mode=mode)


preprocess_input.__doc__ = PREPROCESS_INPUT_DOC.format(
    mode=PREPROCESS_INPUT_MODE_DOC,
    ret="",
    error=PREPROCESS_INPUT_DEFAULT_ERROR_DOC,
)


@keras_export("keras.applications.imagenet_utils.decode_predictions")

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Expand to rank 3/4 with [..., None] or [None, ...]
2. Reshape video data to rank-4 per frame
3. Check ops.shape(images) rank before calling

Example fix

# before
out = ops.image.crop_images(mask, ...)  # (H, W)
# after
out = ops.image.crop_images(mask[..., None], ...)  # (H, W, 1)

if len(ops.shape(images)) not in (3, 4):
    images = backend.convert_to_tensor(images)
    if len(images.shape) == 2:
        images = images[..., None]

def is_crop_safe_rank(t) -> bool:
    return len(t.shape) in (3, 4)

try:
    out = ops.image.crop_images(images, ...)
except ValueError as e:
    if 'rank' in str(e):
        images = normalize_rank(images)
    else:
        raise

When it happens

Trigger: Eager calls to ops.image.crop_images with rank-2 arrays, rank-5 video tensors, or lists converted to wrong-rank tensors.

Common situations: Feeding unlabeled masks; per-frame video processing without reshaping; converting from libraries that drop singleton axes.

Related errors

• weights_path undefined
• The `weights` argument should be either `None` (random initi
• If using `weights="imagenet"` as true, `classes` should be 1
• The number of repeats in `EfficientNet` must be > 0. Receive
• The `weights` argument should be either `None` (random initi

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/6e9e6840dbd87e8d. Report an issue: GitHub.