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

images is expected to be a list of 3d tensors of shape [C, H

Error message

images is expected to be a list of 3d tensors of shape [C, H, W], got {}

What it means

GeneralizedRCNNTransform.forward expects images as a list of individual 3D tensors shaped [C, H, W]. When any tensor has a different dimensionality (commonly a 4D [N, C, H, W] batched tensor), it raises ValueError with the offending shape.

Source

Thrown at pytorch_object_detection/mask_rcnn/network_files/transform.py:439

        _indent = '\n    '
        format_string += "{0}Normalize(mean={1}, std={2})".format(_indent, self.image_mean, self.image_std)
        format_string += "{0}Resize(min_size={1}, max_size={2}, mode='bilinear')".format(_indent, self.min_size,
                                                                                         self.max_size)
        format_string += '\n)'
        return format_string

    def forward(self,
                images,       # type: List[Tensor]
                targets=None  # type: Optional[List[Dict[str, Tensor]]]
                ):
        # type: (...) -> Tuple[ImageList, Optional[List[Dict[str, Tensor]]]]
        images = [img for img in images]
        for i in range(len(images)):
            image = images[i]
            target_index = targets[i] if targets is not None else None

            if image.dim() != 3:
                raise ValueError("images is expected to be a list of 3d tensors "
                                 "of shape [C, H, W], got {}".format(image.shape))
            image = self.normalize(image)  # 对图像进行标准化处理
            image, target_index = self.resize(image, target_index)  # 对图像和对应的bboxes缩放到指定范围
            images[i] = image
            if targets is not None and target_index is not None:
                targets[i] = target_index

        # 记录resize后的图像尺寸
        image_sizes = [img.shape[-2:] for img in images]
        images = self.batch_images(images, self.size_divisible)  # 将images打包成一个batch
        image_sizes_list = torch.jit.annotate(List[Tuple[int, int]], [])

        for image_size in image_sizes:
            assert len(image_size) == 2
            image_sizes_list.append((image_size[0], image_size[1]))

        image_list = ImageList(images, image_sizes_list)
        return image_list, targets

View on GitHub (pinned to 1ec3fe6f37)

Solutions

1. Wrap images in a list of [C,H,W] tensors: model([img1, img2], targets)
2. If you have a batched tensor, split it: images = [t for t in batch]
3. Add a channel dimension to 2D grayscale: img.unsqueeze(0)

Example fix

// before
output = model(batch_images)  # batch_images: [N, C, H, W]
// after
output = model([img for img in batch_images], targets)  # list of [C, H, W] tensors

assert isinstance(images, list) and all(isinstance(t, torch.Tensor) and t.dim() == 3 for t in images), "images must be list of [C,H,W] tensors"
outputs = model(images, targets)

def is_image_list(images):
    return isinstance(images, (list, tuple)) and all(torch.is_tensor(t) and t.dim() == 3 for t in images)

try:
    outputs = model(images, targets)
except ValueError as e:
    if '3d tensors' in str(e):
        images = [img for img in images] if torch.is_tensor(images) else [img.unsqueeze(0) for img in images]
        outputs = model(images, targets)
    else:
        raise

When it happens

Trigger: Passing a batched tensor model(images_tensor) where images_tensor is [N,C,H,W] instead of a list of 3D tensors, or passing a single 2D grayscale image [H,W] unwrapped.

Common situations: Forgetting to index the batch (images[0]); custom collate_fn stacking images into one tensor; feeding PIL-like [H,W] arrays without adding a channel dim.

Understand the failure class

Background: Tensor shape mismatch errors ("must have shape", "expected shape ... got ..."): when tensor dimensions disagree with what an op or layer was told to expect — this error's family across 6 libraries.

Related errors

• images is expected to be a list of 3d tensors of shape [C, H
• expected stages_repeats as list of 3 positive ints
• expected stages_out_channels as list of 5 positive ints
• image: {} isn't RGB mode.
• not find GPU device for training.

AI-assisted analysis of WZMIAOMIAO/deep-learning-for-image-processing@1ec3fe6f37 (2026-08-30). Data as JSON: /api/errors/1e75b2c014760039. Report an issue: GitHub.