RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 92 but got size 60 for tensor number 1 in the list

[PineREN]

Hello, when I use the ppa.yamL to run the model you provided, it will report this error. Am I using it incorrectly? Looking forward to your answer

[03/30 15:50:53 detectron2]: Arguments: Namespace(config_file='configs/sor/ppa.yaml', input='goutu_test', output='output', confidence_threshold=0.4, opts=[])
WARNING [03/30 15:50:53 d2.config.compat]: Config 'configs/sor/ppa.yaml' has no VERSION. Assuming it to be compatible with latest v2.
The checkpoint state_dict contains keys that are not used by the model:
  pixel_mean
  pixel_std
  0%|          | 0/7 [00:00<?, ?it/s]C:\Users\Pine\.conda\envs\SOR\lib\site-packages\torch\functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at  ..\aten\src\ATen\native\TensorShape.cpp:2157.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
  0%|          | 0/7 [00:03<?, ?it/s]
Traceback (most recent call last):
  File "C:\apine\projects\SOR\sor_ppa\vis\demo.py", line 78, in <module>
    predictions, visualized_output = demo.run_on_image(img)
  File "C:\apine\projects\SOR\sor_ppa\vis\predictor.py", line 29, in run_on_image
    predictions = self.predictor(image)
  File "C:\Users\Pine\.conda\envs\SOR\lib\site-packages\detectron2\engine\defaults.py", line 317, in __call__
    predictions = self.model([inputs])[0]
  File "C:\Users\Pine\.conda\envs\SOR\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "C:\Users\Pine\.conda\envs\SOR\lib\site-packages\detectron2\modeling\meta_arch\rcnn.py", line 146, in forward
    return self.inference(batched_inputs)
  File "C:\Users\Pine\.conda\envs\SOR\lib\site-packages\detectron2\modeling\meta_arch\rcnn.py", line 209, in inference
    results, _ = self.roi_heads(images, features, proposals, None)
  File "C:\Users\Pine\.conda\envs\SOR\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "C:\apine\projects\SOR\sor_ppa\centermask\modeling\centermask\center_heads.py", line 458, in forward
    pred_instances = self.forward_with_given_boxes(features, proposals)
  File "C:\apine\projects\SOR\sor_ppa\centermask\modeling\centermask\center_heads.py", line 487, in forward_with_given_boxes
    instances, mask_features, pos = self._forward_mask(features, instances)
  File "C:\apine\projects\SOR\sor_ppa\centermask\modeling\centermask\center_heads.py", line 534, in _forward_mask
    features_.append(torch.cat((feature, y, x), dim=1))
RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 92 but got size 60 for tensor number 1 in the list.

[PineREN]

Hello, is there a requirement for the scale of the input picture ? I found that there was a ratio that did not report errors, so I resized the picture to this ratio, but the result was not ideal. I can solve this by filling in the black edge.

[EricFH]

The size is not fixed, every proposal will get the same size after roi pooling no matter what input shape it is. The scale and ratio is not fixed.