Skip to content

Latest commit

 

History

History

train

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
readme.md
readme.md
 
 
train.sh
train.sh
 
 
train_prior.py
train_prior.py
 
 
transfer.sh
transfer.sh
 
 

readme.md

Train generative priors

Use train.sh to train your generative priors. Before execution, please prepare a list of files like below for your dataset. This dataset contains the complex images that are obtained after phase augmentation.

../abide_2/train/abide_1000000.hdr
../abide_2/train/abide_1000001.hdr
../abide_2/train/abide_1000002.hdr
../abide_2/train/abide_1000003.hdr

The command find could help you to create such a list of files.

folder=/path/to/folder/that/contains/complex/images
pattern="abide_*.hdr"
find $path -name $pattern | sort > /path/to/filelist

Then go to the config file and check following things.

  1. specify where to save the trained models with log_folder
  2. if the variables train_list and test_list are modified in train.sh, please change it in the config file as well
  3. specify the number of GPU available with nr_gpu and gpu_id
  4. please go to spreco for the information on the other configurations