keras-team/keras · error · ValueError

Invalid start_points shape: expected (4,2) for a single…

Error message

Invalid start_points shape: expected (4,2) for a single image or (N,4,2) for a batch. Received shape: {start_points.shape}

What it means

Shape guard in `perspective_transform` (NumPy backend): the source quadrilateral `start_points` must describe 4 corners as a (4, 2) array for a single image or an (N, 4, 2) batch. The guard checks both the ndim (2 or 3) and that the trailing shape equals (4, 2); wrong corner counts, missing coordinate pairs, or an extra/missing batch axis trigger it, with the actual shape shown.

Solutions

1. Provide exactly four (x, y) corner pairs: start_points of shape (4, 2), or (N, 4, 2) for batches
2. Check for dropped coordinate axes: points stored as (4,) or (2, 4) must be reshaped to (4, 2)
3. Ensure batched images (rank 4) get batched start_points and single images (rank 3) get a single quadrilateral

When it happens

Trigger: Thrown at keras/src/backend/numpy/ops/image.py:686 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/5ab6f6e890250351. Report an issue: GitHub.

Appendix: source

Thrown at keras/src/backend/numpy/ops/image.py:686

    start_points = convert_to_tensor(start_points)
    end_points = convert_to_tensor(end_points)

    if interpolation not in AFFINE_TRANSFORM_INTERPOLATIONS:
        raise ValueError(
            "Invalid value for argument `interpolation`. Expected of one "
            f"{AFFINE_TRANSFORM_INTERPOLATIONS}. Received: "
            f"interpolation={interpolation}"
        )

    if len(images.shape) not in (3, 4):
        raise ValueError(
            "Invalid images rank: expected rank 3 (single image) "
            "or rank 4 (batch of images). Received input with shape: "
            f"images.shape={images.shape}"
        )

    if start_points.ndim not in (2, 3) or start_points.shape[-2:] != (4, 2):
        raise ValueError(
            "Invalid start_points shape: expected (4,2) for a single image"
            f" or (N,4,2) for a batch. Received shape: {start_points.shape}"
        )
    if end_points.ndim not in (2, 3) or end_points.shape[-2:] != (4, 2):
        raise ValueError(
            "Invalid end_points shape: expected (4,2) for a single image"
            f" or (N,4,2) for a batch. Received shape: {end_points.shape}"
        )
    if start_points.shape != end_points.shape:
        raise ValueError(
            "start_points and end_points must have the same shape."
            f" Received start_points.shape={start_points.shape}, "
            f"end_points.shape={end_points.shape}"
        )

    input_dtype = images.dtype
    if input_dtype == "float16":
        images = images.astype("float32")

View on GitHub (pinned to 7a34a03db6)