AlexYouXin/Explicit-Shape-Priors

train error

joeyyy-1998 opened this issue · 4 comments

joeyyy-1998 commented

I would very much like to replicate your work.This error occurred when I started training after loading the dataset successfully.
ValueError: Expected target size (1. 128. 160,960），got torch.size([1. 128.151. 96]),I wonder what caused this error,Thank you for your answer.

AlexYouXin commented

Some CT scans do not have the target size 12816096.
please add paddings to the cropped image/GT as shown in our dataloader file.

Minitongg commented

Some CT scans do not have the target size 128_160_96. please add paddings to the cropped image/GT as shown in our dataloader file.

Thank you for your great work! I wonder if the data used for training is cropped or preprocessed of the nnUNet? Thank you for your reply!

AlexYouXin commented

Some CT scans do not have the target size 128_160_96. please add paddings to the cropped image/GT as shown in our dataloader file.

Thank you for your great work! I wonder if the data used for training is cropped or preprocessed of the nnUNet? Thank you for your reply!

Here the vertebral data used for training is preprocessed with only the z-score normalization strategy by nnUNet, without resampling or cropping strategies.

Minitongg commented

Some CT scans do not have the target size 128_160_96. please add paddings to the cropped image/GT as shown in our dataloader file.

Thank you for your great work! I wonder if the data used for training is cropped or preprocessed of the nnUNet? Thank you for your reply!

Here the vertebral data used for training is preprocessed with only the z-score normalization strategy by nnUNet, without resampling or cropping strategies.

I got it ! Thank you for your generous answers.