search

yhygao / CBIM-Medical-Image-Segmentation

A PyTorch framework for medical image segmentation
Apache License 2.0
260 stars 46 forks source link

类型转换问题，tensor_lab = torch.from_numpy(lab).long() #2

Closed ruifenggong closed 2 years ago

ruifenggong commented 2 years ago

https://github.com/yhygao/CBIM-Medical-Image-Segmentation/blob/2f4408900378bf6600673bbe10dbcbf60d646e35/training/dataset/dim3/dataset_acdc.py#L112

In dataloading processing， error accurs intensor_lab = torch.from_numpy(lab).long() ,

But it works when I add a line lab = lab/1.0 before that.It is a bug here?

yhygao commented 2 years ago

There is no error in my experiment. Can you show the error code? It might be because your 'lab' is already a long type, so it can't be cast as a long type. But it's actually not a big problem, as this line aims at converting label data to integer for loss computation. It's fine if you fix the error by doing so.

ruifenggong commented 2 years ago

There is no error in my experiment. Can you show the error code? It might be because your 'lab' is already a long type, so it can't be cast as a long type. But it's actually not a big problem, as this line aims at converting label data to integer for loss computation. It's fine if you fix the error by doing so.

The log is here：

Traceback (most recent call last):
  File "/home/xxx/CBIM-Medical-Image-Segmentation/train.py", line 248, in <module>
    best_Dice, best_HD, best_ASD = train_net(net, args, ema_net, fold_idx=i)
  File "/home/xxx/CBIM-Medical-Image-Segmentation/train.py", line 37, in train_net
    trainset = get_dataset(args, mode='train', fold_idx=fold_idx)
  File "/home/xxx/CBIM-Medical-Image-Segmentation/training/dataset/utils.py", line 20, in get_dataset
    return CMRDataset(args, mode=mode)
  File "/home/xxx/CBIM-Medical-Image-Segmentation/training/dataset/dim3/dataset_bileduck.py", line 57, in __init__
    img, lab = self.preprocess(itk_img, itk_lab, img_name, lab_name)
  File "/home/xxx/CBIM-Medical-Image-Segmentation/training/dataset/dim3/dataset_bileduck.py", line 125, in preprocess
    tensor_lab = torch.from_numpy(lab)
TypeError: can't convert np.ndarray of type numpy.uint16. The only supported types are: float64, float32, float16, complex64, complex128, int64, int32, int16, int8, uint8, and bool.
yhygao commented 2 years ago

The error code indicates that the data type of 'lab' is np.uint16. torch.from_numpy() doesn't support this data type. You can cast lab to uint8 first: lab = lab.astype(np.uint8)

This error depends on your data, but I'll add a data type check and cast to avoid this error. Thanks!