Error when using illu_orient in DeStripe

[dpshepherd]

Hello,

Thanks for this great software. We are trying it on some single illumination SPIM data using the leonardo_toolset version installed pypi. Following the tutorial, we were trying to use the illumination orientation as follows:

    recon = model.train(
        is_vertical = True,
        display = False,
        display_angle_orientation=False,
        angle_offset = [0],
        x = test_plane,
        illu_orient="bottom"
    )

We get the following error (full traceback):

[Leonardo-DeStripe] Backend: PyTorch | Device: cuda
Start DeStripe...

C:\Users\q2ilab\miniconda3\envs\leonardo\lib\site-packages\torch\nn\modules\module.py:1341: UserWarning: Complex modules are a new feature under active development whose design may change, and some modules might not work as expected when using complex tensors as parameters or buffers. Please file an issue at https://github.com/pytorch/pytorch/issues/new?template=bug-report.yml if a complex module does not work as expected.
  warnings.warn(
bottom
  0%|                                                                                                                                                                                                              | 0/1000 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "c:\Users\q2ilab\Documents\github\curved_aslm\test_leonardo.py", line 39, in <module>
    main()
  File "c:\Users\q2ilab\Documents\github\curved_aslm\test_leonardo.py", line 21, in main
    recon = model.train(
  File "C:\Users\q2ilab\miniconda3\envs\leonardo\lib\site-packages\leonardo_toolset\destripe\core.py", line 784, in train
    out = self.train_on_full_arr(
  File "C:\Users\q2ilab\miniconda3\envs\leonardo\lib\site-packages\leonardo_toolset\destripe\core.py", line 542, in train_on_full_arr
    Y, target = DeStripe.train_on_one_slice(
  File "C:\Users\q2ilab\miniconda3\envs\leonardo\lib\site-packages\leonardo_toolset\destripe\core.py", line 317, in train_on_one_slice
    Y = post_process_module(
  File "C:\Users\q2ilab\miniconda3\envs\leonardo\lib\site-packages\leonardo_toolset\destripe\post_processing.py", line 564, in post_process_module
    hX = linear_propagation(
  File "C:\Users\q2ilab\miniconda3\envs\leonardo\lib\site-packages\leonardo_toolset\destripe\post_processing.py", line 440, in linear_propagation
    b_new_flip = train_post_process_module(
  File "C:\Users\q2ilab\miniconda3\envs\leonardo\lib\site-packages\leonardo_toolset\destripe\post_processing.py", line 289, in train_post_process_module
    b_new, decay, weight, b_new2 = model(
  File "C:\Users\q2ilab\miniconda3\envs\leonardo\lib\site-packages\torch\nn\modules\module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "C:\Users\q2ilab\miniconda3\envs\leonardo\lib\site-packages\torch\nn\modules\module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
  File "C:\Users\q2ilab\miniconda3\envs\leonardo\lib\site-packages\leonardo_toolset\destripe\post_processing.py", line 74, in forward
    b_adpt = b * fusion_mask + b_old * (1 - fusion_mask)
RuntimeError: The size of tensor a (285) must match the size of tensor b (234) at non-singleton dimension 2

Any suggestions on how to properly use the illumination orientation? If we do not use the illu_orient parameter, everything works fine.

Thanks!

[jxchen01]

@dpshepherd thanks for testing it out. could you let us know the size of your input images? We can try to reproduce the error on our end. @yuliu96 can you take a look?

[yuliu96]

hi @dpshepherd thanks for trying out destripe? could you share the shape of the test_palne so that i can reproduce and fix it?