WZMIAOMIAO/deep-learning-for-image-processing · error · Exception

The input image should np.float32 in the range [0, 1]

Error message

The input image should np.float32 in the range [0, 1]

What it means

grad_cam's show_cam_on_image overlays a heatmap onto img, which must be a float32 array normalized to [0, 1]. It detects values > 1 (typically a uint8 0-255 image) and raises rather than producing a washed-out overlay. The check guards the additive blending cam = heatmap + img.

Source

Thrown at pytorch_classification/grad_cam/utils.py:198

                      use_rgb: bool = False,
                      colormap: int = cv2.COLORMAP_JET) -> np.ndarray:
    """ This function overlays the cam mask on the image as an heatmap.
    By default the heatmap is in BGR format.

    :param img: The base image in RGB or BGR format.
    :param mask: The cam mask.
    :param use_rgb: Whether to use an RGB or BGR heatmap, this should be set to True if 'img' is in RGB format.
    :param colormap: The OpenCV colormap to be used.
    :returns: The default image with the cam overlay.
    """

    heatmap = cv2.applyColorMap(np.uint8(255 * mask), colormap)
    if use_rgb:
        heatmap = cv2.cvtColor(heatmap, cv2.COLOR_BGR2RGB)
    heatmap = np.float32(heatmap) / 255

    if np.max(img) > 1:
        raise Exception(
            "The input image should np.float32 in the range [0, 1]")

    cam = heatmap + img
    cam = cam / np.max(cam)
    return np.uint8(255 * cam)


def center_crop_img(img: np.ndarray, size: int):
    h, w, c = img.shape

    if w == h == size:
        return img

    if w < h:
        ratio = size / w
        new_w = size
        new_h = int(h * ratio)
    else:

View on GitHub (pinned to 1ec3fe6f37)

Solutions

1. Normalize before the call: img = np.float32(img) / 255 (and convert BGR->RGB if needed)
2. Use cv2.cvtColor + astype(np.float32)/255.0 immediately after cv2.imread
3. Clamp/verify range with assert np.max(img) <= 1 in your preprocessing

Example fix

// before
rgb_img = cv2.imread(path)
cv2.waitKey()
show_cam_on_image(rgb_img, grayscale_cam)
// after
rgb_img = cv2.imread(path)[:, :, ::-1]
rgb_img = np.float32(rgb_img) / 255
show_cam_on_image(rgb_img, grayscale_cam)

img = cv2.imread(path)[:, :, ::-1]
assert img.dtype == np.uint8
img = np.float32(img) / 255.0
assert 0.0 <= img.min() and img.max() <= 1.0

def is_unit_float_image(img: 'np.ndarray') -> bool:
    return img.dtype == np.float32 and img.max() <= 1.0 and img.min() >= 0.0

try:
    visualization = show_cam_on_image(img, grayscale_cam)
except Exception as e:
    if 'np.float32 in the range [0, 1]' in str(e):
        img = np.float32(img) / 255.0
        visualization = show_cam_on_image(img, grayscale_cam)
    else:
        raise

When it happens

Trigger: Calling show_cam_on_image(img, mask) with img read by cv2.imread or kept as np.uint8 in [0, 255]; also fires if img is float32 but scaled beyond 1.0.

Common situations: Forgetting the standard preprocessing img = np.float32(img) / 255 after cv2.imread; passing a PIL image converted via np.array without scaling; mixing BGR uint8 OpenCV reads with CAM utilities.

Related errors

• not support data format '{self.data_format}'
• image: {} isn't RGB mode.
• Transformer input dimension should be divisible by head dime
• image: {} isn't RGB mode.
• Embedding dim must be divisible by number of heads in {}. Go

AI-assisted analysis of WZMIAOMIAO/deep-learning-for-image-processing@1ec3fe6f37 (2026-08-30). Data as JSON: /api/errors/1f30622fb01915e3. Report an issue: GitHub.