This is a refine version of DeepLung repo with dependecies:
- python 3.7
- pytorch 1.10
Reference:
- Download the luna16 data and put all them in folder
luna16
- change
path
variable in fileconfig_training.py
andconfig_training9.py
to your data folder. (changef
variable inrun_training.sh
to choose another config file) - run
prepare.py
to preproess all data.
- use the cmd
sh run_training.sh
- torch version problem
- data path
- numpy version problem (1.21.5 with problem of np.asarray(metrics))
- fix the error loading file with xlrd : /data/yangqinzhu/ctLung/DeepLung-master/nodcls/data/list3.2.csv
- change
.dcm
file name, such as1-001.dcm
to0000001.dcm
- run
./nodcls/data/extclsshpinfo.py
- run
/data/yangqinzhu/ctLung/DeepLung-master/nodcls/data/dataconverter.py
to get fileannotationdetclsconv_v3.csv
- run
/data/yangqinzhu/ctLung/DeepLung-master/nodcls/data/nodclsgbt.py
to generate folder/luna16/cls/crop_v3/
- run
main_nodcls.py
to train the based classify model.(/data/yangqinzhu/ctLung/DeepLung-master/checkpoint-9/ckpt.t7
) - run
bash run_test_getTrainData.sh
to get pbb train data before runing filedet2cls.py
. - run
det2cls.py
to train classify. - run
testdet2cls.py
.
you can change the path as you need, I have listed them it front of the function in each python file