huggingface/transformers · error · ValueError

Unsupported input type {type(bboxes_corners)}

Error message

Unsupported input type {type(bboxes_corners)}

What it means

corners_to_center_format is the inverse conversion with the same dispatch rule: only torch tensors and numpy arrays are accepted; lists or other tensor types raise ValueError. It converts (top_left_x, top_left_y, bottom_right_x, bottom_right_y) to (center_x, center_y, width, height).

Source

Thrown at src/transformers/image_transforms.py:608

    return bboxes_center


def corners_to_center_format(bboxes_corners: TensorType) -> TensorType:
    """
    Converts bounding boxes from corners format to center format.

    corners format: contains the coordinates for the top-left and bottom-right corners of the box
        (top_left_x, top_left_y, bottom_right_x, bottom_right_y)
    center format: contains the coordinate for the center of the box and its the width, height dimensions
        (center_x, center_y, width, height)
    """
    # Inverse function accepts different input types so implemented here too
    if is_torch_tensor(bboxes_corners):
        return _corners_to_center_format_torch(bboxes_corners)
    elif isinstance(bboxes_corners, np.ndarray):
        return _corners_to_center_format_numpy(bboxes_corners)

    raise ValueError(f"Unsupported input type {type(bboxes_corners)}")


def safe_squeeze(
    tensor: Union[np.ndarray, "torch.Tensor"], axis: int | None = None
) -> Union[np.ndarray, "torch.Tensor"]:
    """
    Squeezes a tensor, but only if the axis specified has dim 1.
    """
    if axis is None:
        return tensor.squeeze()

    try:
        return tensor.squeeze(axis=axis)
    except ValueError:
        return tensor


# 2 functions below copied from https://github.com/cocodataset/panopticapi/blob/master/panopticapi/utils.py

View on GitHub (pinned to a597f97485)

Solutions

1. Wrap with np.asarray(boxes) (or torch.tensor(boxes) if in a torch pipeline).
2. Keep the tensor type consistent with the rest of the pipeline.
3. Validate boxes.shape[-1] == 4 before calling.

Example fix

# before
centers = corners_to_center_format([[0, 0, 100, 100]])  # raises

# after
import numpy as np
centers = corners_to_center_format(np.array([[0, 0, 100, 100]]))

import numpy as np
if not (is_torch_tensor(bboxes_corners) or isinstance(bboxes_corners, np.ndarray)):
    bboxes_corners = np.asarray(bboxes_corners, dtype=float)
assert bboxes_corners.shape[-1] == 4

def is_supported_boxes(x) -> bool:
    import numpy as np
    from transformers.utils import is_torch_tensor
    return is_torch_tensor(x) or isinstance(x, np.ndarray)

When it happens

Trigger: corners_to_center_format(boxes_list) where boxes_list is a Python list/tuple of corner coordinates, or a non-torch tensor framework object.

Common situations: Post-processing detection outputs into COCO-style annotations with plain-list boxes, or feeding API/JSON payloads (which deserialize to lists) into geometry helpers.

Related errors

• Unsupported input type {type(bboxes_center)}
• `axis_value` for `HQQ` backend has to be one of [`0`, `1`] b
• You can construct a Cache either from a list `layers` of all
• You should provide exactly one of `layers` or `layer_class_t
• Cannot call `update_conv_state` on a non-LinearAttention lay

AI-assisted analysis of huggingface/transformers@a597f97485 (2026-08-14). Data as JSON: /api/errors/df4267796551a983. Report an issue: GitHub.