roboflow/supervision · error · ValueError

Unsupported interpolation mode: {interpolation}

Error message

Unsupported interpolation mode: {interpolation}

What it means

The fallback resize implements only two interpolation modes: INTER_NEAREST (index-based gather) and INTER_LINEAR (Pillow bilinear or NumPy affine sampling). Any other cv2 interpolation constant (INTER_CUBIC, INTER_AREA, INTER_LANCZOS4, etc.) is rejected rather than silently approximated.

Source

Thrown at src/supervision/_cv2/_image.py:157

    source_height, source_width = src.shape[:2]
    width, height = dsize if dsize is not None else (0, 0)
    if width == 0 or height == 0:
        width = round(source_width * fx)
        height = round(source_height * fy)
    if min(width, height, source_width, source_height) <= 0:
        raise ValueError("Resize dimensions must be positive")

    if interpolation == _INTER_NEAREST:
        y_indices = np.minimum(
            (np.arange(height) * source_height // height), source_height - 1
        )
        x_indices = np.minimum(
            (np.arange(width) * source_width // width), source_width - 1
        )
        return np.ascontiguousarray(src[y_indices[:, np.newaxis], x_indices])

    if interpolation != _INTER_LINEAR:
        raise ValueError(f"Unsupported interpolation mode: {interpolation}")

    if src.dtype == np.uint8 and (
        src.ndim == 2 or (src.ndim == 3 and src.shape[2] == 3)
    ):
        from PIL import Image

        size = (width, height)
        image = Image.fromarray(src)
        if width >= source_width and height >= source_height:
            resized = image.resize(size, resample=Image.Resampling.BILINEAR)
        else:
            # Affine sampling keeps Pillow from widening its bilinear kernel
            # during reduction and maps pixel centers like INTER_LINEAR.
            resized = image.transform(
                size,
                Image.Transform.AFFINE,
                (source_width / width, 0, 0, 0, source_height / height, 0),
                resample=Image.Resampling.BILINEAR,

View on GitHub (pinned to 7f254d9784)

Solutions

1. Use cv2.INTER_LINEAR or cv2.INTER_NEAREST, the two supported modes.
2. Install opencv-python (or opencv-python-headless) so Supervision delegates to real cv2 and all modes work.
3. If you need another kernel, pre-resize with PIL/scipy yourself and skip the fallback path.

Example fix

# before
small = cv2.resize(frame, (w, h), interpolation=cv2.INTER_AREA)

# after (fallback-compatible)
small = cv2.resize(frame, (w, h), interpolation=cv2.INTER_LINEAR)

SUPPORTED = {cv2.INTER_NEAREST, cv2.INTER_LINEAR}
interp = interp if interp in SUPPORTED else cv2.INTER_LINEAR
resized = cv2.resize(frame, (w, h), interpolation=interp)

When it happens

Trigger: Calling cv2.resize(..., interpolation=cv2.INTER_AREA) or INTER_CUBIC/INTER_LANCZOS4/INTER_NEAREST_EXACT etc. while running on the Supervision OpenCV fallback.

Common situations: Code written against real OpenCV that uses INTER_AREA for downscaling (a common high-quality shrink idiom) or INTER_CUBIC for upscaling, then executed in an environment where opencv-python is not installed (slim Docker images, CI without cv2).

Related errors

• Resize dimensions must be positive
• PyAV fallback supports file paths, not webcam device indexes
• addWeighted inputs must have equal shapes
• Mean mask must match the image height and width
• Unsupported flip code: {flip_code}

AI-assisted analysis of roboflow/supervision@7f254d9784 (2026-08-15). Data as JSON: /api/errors/9e31c0ed9d98aeac. Report an issue: GitHub.