UnetTumorSegment

setup vm

Download model and data

1. Download cuda and set the sourcing in the bashrc

https://developer.nvidia.com/cuda-11-7-0-download-archive?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=20.04&target_type=deb_local
export PATH="/usr/local/cuda-XX.X/bin:$PATH"
nvcc --version
sudo prime-select nvidia
prime-select query

2. clone github repo (setup ssh keys):

connect the ssh keys
git clone [email protected]:BadrEssabri/UnetTumorSegment.git

2. Download required packages: dont use pip install -r requirements.txt

pip install torch torchvision torchaudio
pip install h5py matplotlib tqdm 

3. Download dataset from Google Drive (train_brain and val_brain):

pip install gdown
cd UnetTumorSegment/data/train_brain
gdown --id 1_ePoCdcF6qOVp5SDPRffG7nwMLjvUaJm
unzip train_brain.zip
rm -rf train_brain.zip
gdown --id 13KCXTn1_fD9f-oLq9FIf5e8LeaUgJjot
unzip val_brain.zip
rm -rf val_brain.zip

Test the program

1. Check all the paths in the python program
2. Run the program

python3 ./train.py

3. Check if no errors occur

Run the program as background process

To run the program, even if you closed the VM, you need to run the script as a background process and save the outputs in a text file.

1. Get location of python version

which python

2. Start the background process

nohup <absolute path to anaconda python> -u <absolute path to python script> > outputfile.txt &

3. Check if the outputfile.txt contains the output of the program, if not something went wrong

cat outputfile.txt

4. Check if the process is running in the background, you should see multiple running process

ps aux | grep <Name of running script>

5. If you want to check the status of your program at a later time

cat outputfile.txt

Download results

If the program is finished, download the results (weights of the model)