Function resize_numpy_image

Adapter/utils.py:24–34

(image, max_resolution=1024 * 1024, resize_short_edge=None)

        raise ValueError(f"{model_class} is not supported.")

def resize_numpy_image(image, max_resolution=1024 * 1024, resize_short_edge=None):
    h, w = image.shape[:2]
    if resize_short_edge is not None:
        k = resize_short_edge / min(h, w)
    else:
        k = max_resolution / (h * w)
        k = k**0.5
    h = int(np.round(h * k / 64)) * 64
    w = int(np.round(w * k / 64)) * 64
    image = cv2.resize(image, (w, h), interpolation=cv2.INTER_LANCZOS4)
    return image