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

Expected target boxes to be a tensorof shape [N, 4], got {:}

Error message

Expected target boxes to be a tensorof shape [N, 4], got {:}.

What it means

During target validation in FasterRCNN.forward, if target['boxes'] is a torch.Tensor it must be 2-D with last dimension 4 ([N,4] xyxy boxes); otherwise this ValueError is raised with the actual shape. It enforces the box tensor contract before boxes flow into the RPN/ROI heads.

Source

Thrown at pytorch_object_detection/faster_rcnn/network_files/faster_rcnn_framework.py:68

            targets (list[Dict[Tensor]]): ground-truth boxes present in the image (optional)

        Returns:
            result (list[BoxList] or dict[Tensor]): the output from the model.
                During training, it returns a dict[Tensor] which contains the losses.
                During testing, it returns list[BoxList] contains additional fields
                like `scores`, `labels` and `mask` (for Mask R-CNN models).

        """
        if self.training and targets is None:
            raise ValueError("In training mode, targets should be passed")

        if self.training:
            assert targets is not None
            for target in targets:         # 进一步判断传入的target的boxes参数是否符合规定
                boxes = target["boxes"]
                if isinstance(boxes, torch.Tensor):
                    if len(boxes.shape) != 2 or boxes.shape[-1] != 4:
                        raise ValueError("Expected target boxes to be a tensor"
                                         "of shape [N, 4], got {:}.".format(
                                          boxes.shape))
                else:
                    raise ValueError("Expected target boxes to be of type "
                                     "Tensor, got {:}.".format(type(boxes)))

        original_image_sizes = torch.jit.annotate(List[Tuple[int, int]], [])
        for img in images:
            val = img.shape[-2:]
            assert len(val) == 2  # 防止输入的是个一维向量
            original_image_sizes.append((val[0], val[1]))
        # original_image_sizes = [img.shape[-2:] for img in images]

        images, targets = self.transform(images, targets)  # 对图像进行预处理

        # print(images.tensors.shape)
        features = self.backbone(images.tensors)  # 将图像输入backbone得到特征图
        if isinstance(features, torch.Tensor):  # 若只在一层特征层上预测，将feature放入有序字典中，并编号为‘0’

View on GitHub (pinned to 1ec3fe6f37)

Solutions

1. Reshape boxes to [N,4]: boxes = boxes.view(-1, 4) or torch.as_tensor(boxes).reshape(-1, 4).
2. For a single box, wrap it: boxes = boxes.unsqueeze(0).
3. Keep labels in target['labels'], not inside target['boxes'].
4. Verify each target before training: check boxes.ndim == 2 and boxes.shape[1] == 4.
5. Print boxes.shape for the offending target to see the actual layout.

Example fix

# before
target = {'boxes': torch.tensor([10., 20., 110., 120.]), 'labels': torch.tensor([1])}
# after
target = {'boxes': torch.tensor([[10., 20., 110., 120.]]), 'labels': torch.tensor([1])}  # shape [1,4]

for t in targets:
    b = t['boxes']
    assert isinstance(b, torch.Tensor) and b.ndim == 2 and b.shape[-1] == 4, f"bad boxes shape {b.shape if hasattr(b,'shape') else b}"

def is_valid_box_tensor(b) -> bool:
    return isinstance(b, torch.Tensor) and b.ndim == 2 and b.shape[-1] == 4

try:
    loss_dict = model(images, targets)
except ValueError as e:
    if 'shape [N, 4]' in str(e):
        targets = [{'boxes': t['boxes'].view(-1, 4), 'labels': t['labels']} for t in targets]
        loss_dict = model(images, targets)
    else:
        raise

When it happens

Trigger: Passing boxes with wrong shape — e.g. shape [N] (flattened), [4] (single box unbatched), [N,5] (with extra column), or a list of per-coordinate values stored as a tensor of wrong rank — in training targets.

Common situations: Custom datasets building targets incorrectly, forgetting torch.stack/torch.as_tensor around per-box rows, concatenating labels into the boxes tensor, or loading boxes as normalized [0,1] values in a different layout.

Related errors

• Expected target boxes to be of type Tensor, got {:}.
• In training mode, targets should be passed
• images is expected to be a list of 3d tensors of shape [C, H
• No ground-truth boxes available for one of the images during
• No proposal boxes available for one of the images during tra

AI-assisted analysis of WZMIAOMIAO/deep-learning-for-image-processing@1ec3fe6f37 (2026-08-30). Data as JSON: /api/errors/13e44162c17ae5bc. Report an issue: GitHub.