Comments (7)
Hi again,
today I tried with an RTX 2080 (8GB) with a similar result:
RuntimeError: CUDA out of memory. Tried to allocate 2.06 GiB (GPU 0; 8.00 GiB total capacity; 4.21 GiB already allocated; 0 bytes free; 6.28 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
I think in the paper a GPU with 12GB was used?
from quicknat_pytorch.
Hi,
Seems you have run the code. But I cannot understand how the 'converth5.py' works especially the labels. Could you give me a hint? I think the author of this paper will never answer any questions...
from quicknat_pytorch.
The paper mentioned in 2.3: "We use a constant weight decay of 0.0001. Batch size is set to 4, limited by the 12GB RAM of the NVIDIA TITAN X Pascal GPU."
from quicknat_pytorch.
The paper mentioned that they use FreeSurfer to handle IXI Dataset. But when I started to learn FreeSurfer, I found that FreeSurfer will not give a single "Auxiliary label". The output of FreeSurfer contains many files including a folder named "label". And I think the folder is not the "label" in "convert_h5.py". So, how to start training?
from quicknat_pytorch.
Hi,
I never executed training, only segmentation.
I preprocessed my niftis with
mri_convert --conform <input.nii> <output.nii>
Regarding executing QuickNAT, my steps are as follows:
-
install quicknat
- install cuda 11.3
- install python 3.7.9 (x64)
- make sure the correct python is in path
- create a folder D:/quicknat_test
- copy the folder called after the github commit hash to D:/quicknat_test/src
- edit "settings_eval.ini" - change the following:
data_dir = "D:/quicknat_test/nifti/process"
directory_struct = "Linear"
estimate_uncertainty = "True" - start cmd with admin permissions
- install virtualenv for python (if not already done)
pip install virtualenv - create a virtual environment
virtualenv D:/quicknat_test/env - activate the virtual environment
D:\quicknat_test\env\Scripts\activate.bat - go to the folder D:\quicknat_test\src\4e4e97e912b9f75f9c299065922009da737c4ef9
- install correct torch
pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113 - install the rest of the dependencies
python -m pip install -r requirements.txt
-
execute quicknat
- copy the preprocessed nifti files into D:/quicknat_test/nifti/process
- edit "test_list.txt" and enter the filenames in the data_dir you want to process
- start cmd with admin permissions
- activate the virtual environment
D:\quicknat_test\env\Scripts\activate.bat - go to the src folder:
cd D:\quicknat_test\src\4e4e97e912b9f75f9c299065922009da737c4ef9 - start processing
python run.py --mode=eval_bulk - results under:
D:\quicknat_test\src\4e4e97e912b9f75f9c299065922009da737c4ef9\ixi_test_seg\one_view
from quicknat_pytorch.
Thanks! I'll try it!
from quicknat_pytorch.
The paper mentioned that they use FreeSurfer to handle IXI Dataset. But when I started to learn FreeSurfer, I found that FreeSurfer will not give a single "Auxiliary label". The output of FreeSurfer contains many files including a folder named "label". And I think the folder is not the "label" in "convert_h5.py". So, how to start training?
Hi, for training using FreeSurfer segmentations you can use the segmentation file: mri/aseg.mgz which contains the segmentation of subcortical structures used in QuickNat, and the mri volume: mri/orig.mgz. In utils/preprocessor.py is a function remap_labels that shows which of the classes were used.
from quicknat_pytorch.
Related Issues (20)
- ModuleNotFoundError: No module named 'utils' HOT 1
- Re-train the model with new data: 3 view aggregation HOT 1
- I can't find pretrained model HOT 4
- Why use PReLU in DenseBlock? HOT 1
- Predicting input nii files? HOT 11
- Fixed - Installing Requirements Error: squeeze_and_excite HOT 2
- "cpu" or cpu not supported as device in settings_eval.ini HOT 2
- "EOL while scanning string literal" for .nii file
- Large Binaries in Repo
- 'phase' not used in cm_per_epoch() HOT 1
- Question about `mri_convert` HOT 2
- training example HOT 1
- Run with new data HOT 1
- docs: information about datasets used for pretrained models
- Seems like a bug while training my own module... HOT 1
- what did squeeze_and_excitation module do?
- Code documentation HOT 1
- Pypi new package version release HOT 1
- Checkpoint settings reload HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from quicknat_pytorch.