confusezius/unet-lits-2d-pipeline

Liver Lesion Segmentation with 2D Unets

unet-lits-2d-pipeline's Issues

GPU use

I want to run the program with two GPUs. I fixed os.environ['CUDA_LAUNCH_BLOCKING'] = "0, 1" in Train_Networks but it is still unsuccessful. I would like to ask how to do this.

Time limit exceeded on lits challenge

hello,
yesterday i made a submission on codalab but after running a couple of hours i was surprised by a weird ''time limit exceeded" error do you have any idea about the origin of this error ? thank you

Boundary_weightmap

Thank you for your sharing. I have some quesions ：

1.  “   boundary_weightmap_lesion = 0.75*np.clip(np.exp(-0.02*(ndi.morphology.distance_transform_edt(1-total_border))),weight,None)+0.25*lesion_mask
   

   boundary_weightmap_liver = 0.65boundary_weightmap_liver + 0.35boundary_weightmap_lesion”
   Q：
   (1) Did boundary_weightmap formula come from ‘ weight_map = wc+w0 * np.exp(-1/2 * ((d1+d2)/sigma) ** 2) ’？
   (2) How to determine the coefficients of the weight formula to be 0.75 and 0.25，0.65 and 0.35？
   (3) How is this boundary_weightmap sent to the network? to which part of the network？

2." if opt.Training['data']=='lesion': vol_info['ref_mask_info']= pd.read_csv(opt.Paths['Training_Path']+'/Assign_2D_LiverMasks.csv', header=0)"
Q：What does this code mean？Why 'data' is'lesion', but Assign_2D_LiverMasks.csv is given to ref_mask_info？

Looking forward to your answer, thank you very much！

pretrained weight

Thank you very much for your sharing! I have a simple task using the segmented liver for postpropressing with my own datasets, so if you can provide the weight pretrained in the dataset ? Thanks again!

train with self-made dataset

Thanks for sharing the great work. I am a student working on similar project and I'm still learning the techniques, I wonder how could I adjust the data preparation code to use self-made dataset including binary masks stored in png and sliced volume stored in mat instead of LITS database, thanks in advance for your help.

about your result on CodaLab

Hi，

Thanks for your great work, could you share you result on LITS17 Challenge, thanks!

I got some error when training

When running

python Train_Networks/Train_Networks.py --search_setup Small_UNet_Liver.txt --no_date

i got this errror, and dont know how to fix it.

Setup Iteration... :   0%|                                                                    | 0/1 [00:00<?, ?it/s]│
Traceback (most recent call last):                                                                                 
  File "Train_Networks/Train_Networks.py", line 232, in <module>                                                    
    main(training_setup)                                                                                           
  File "Train_Networks/Train_Networks.py", line 114, in main                                                        
    train_dataset, val_dataset = Data.Generate_Required_Datasets(opt)                                               
  File "/home/cc/jupyter/unet-lits-2d-pipeline/Train_Networks/../Utilities/PyTorch_Datasets.py", line 23, in Generat
e_Required_Datasets                                                                                                 
    available_volumes = sorted(list(set(np.array(vol_info['volume_slice_info']['Volume']))), key=lambda x: int(x.spl
it('-')[-1]))                                                                                                      
  File "/home/cc/jupyter/unet-lits-2d-pipeline/Train_Networks/../Utilities/PyTorch_Datasets.py", line 23, in <lambda                                                                                   
    available_volumes = sorted(list(set(np.array(vol_info['volume_slice_info']['Volume']))), key=lambda x: int(x.spl
it('-')[-1]))                                                                                                       
ValueError: invalid literal for int() with base 10: 'Volume'  

Could u please help me?

creating weight map

I have some trouble creating weight map using Convert_NibVolumes_to_Slices.py (not the 3D version since my data are already slices). The program outputs:
Reloaded modules: network_zoo, Network_Utilities, General_Utilities, PyTorch_Datasets, Function_Library
0%| | 0/1 [00:00<?, ?it/s]
Without weightmap or csv file. I noticed that on L178 of that file explaining that txt file are read, I wonder which txt file should I adjust? Thanks :)

RuntimeError

RuntimeError: CUDA out of the memory. Tried to allocate 360.00 MiB

GPU using

In the program, I want to run the program with two GPUs. I fixed os.environ['CUDA_LAUNCH_BLOCKING'] = "0, 1" in Train_Networks but it is still unsuccessful. I would like to ask how to do this.

Execution Error

May I know what is this error about

0% 0/4165 [00:00<?, ?it/s]
train_iter 4165
0% 0/4165 [00:00<?, ?it/s]
(#26453082) Training [lr=[3.e-05]]: 0% 0/50 [00:00<?, ?it/s]
Setup Iteration... : 0% 0/1 [00:05<?, ?it/s]
Traceback (most recent call last):
File "/content/drive/MyDrive/My Drive/Tversky Loss/Train_Networks/Train_Networks.py", line 235, in
main(training_setup)
File "/content/drive/MyDrive/My Drive/Tversky Loss/Train_Networks/Train_Networks.py", line 137, in main
flib.trainer([network,optimizer], train_data_loader, [base_loss_func, aux_loss_func], opt, Metrics, epoch)
File "/content/drive/MyDrive/My Drive/Tversky Loss/Train_Networks/Function_Library.py", line 36, in trainer
for slice_idx, file_dict in enumerate(train_data_iter):
File "/usr/local/lib/python3.7/dist-packages/tqdm/std.py", line 1180, in iter
for obj in iterable:
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py", line 521, in next
data = self._next_data()
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py", line 1203, in _next_data
return self._process_data(data)
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py", line 1229, in _process_data
data.reraise()
File "/usr/local/lib/python3.7/dist-packages/torch/_utils.py", line 434, in reraise
raise exception
ValueError: Caught ValueError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/_utils/worker.py", line 287, in _worker_loop
data = fetcher.fetch(index)
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/_utils/fetch.py", line 49, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/_utils/fetch.py", line 49, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/content/drive/MyDrive/My Drive/Tversky Loss/Utilities/PyTorch_Datasets.py", line 161, in getitem
crops_for_picked_batch = gu.get_crops_per_batch(files_to_crop, crop_mask, crop_size=self.pars.Training['crop_size'], seed=self.rng.randint(0,1e8))
File "/content/drive/MyDrive/My Drive/Tversky Loss/Utilities/General_Utilities.py", line 307, in get_crops_per_batch
crop_idx = [rng.randint(crop_size[i]//2-1,np.array(batch_to_crop[0].shape[i+1])-crop_size[i]//2-1) for i in range(batch_to_crop[0].ndim-1)]
File "/content/drive/MyDrive/My Drive/Tversky Loss/Utilities/General_Utilities.py", line 307, in
crop_idx = [rng.randint(crop_size[i]//2-1,np.array(batch_to_crop[0].shape[i+1])-crop_size[i]//2-1) for i in range(batch_to_crop[0].ndim-1)]
File "mtrand.pyx", line 746, in numpy.random.mtrand.RandomState.randint
File "_bounded_integers.pyx", line 1254, in numpy.random._bounded_integers._rand_int64
ValueError: low >= high