How Do I Run a Simple GAN without any conditioning?

[enochkan]

How Do I Run a Simple GAN without any conditioning?

I have downloaded the source code and ran net_gan, which uses gan_application.py and obtained the following error:

CRITICAL:niftynet: Reader requires task input keywords ('conditioning',), but not exist in the config file.
Available task keywords: ['noise_size', 'image', 'n_interpolations', 'conditioning', 'name']
Traceback (most recent call last):
File "net_gan.py", line 8, in
sys.exit(main())
File "/Users/kan7792/NiftyNet/niftynet/init.py", line 141, in main
app_driver.initialise_application(system_param, input_data_param)
File "/Users/kan7792/NiftyNet/niftynet/engine/application_driver.py", line 167, in initialise_application
data_param, app_param, self.data_partitioner)
File "/Users/kan7792/NiftyNet/niftynet/application/gan_application.py", line 69, in initialise_dataset_loader
data_param, task_param, file_list) for file_list in file_lists]
File "/Users/kan7792/NiftyNet/niftynet/io/image_reader.py", line 142, in initialise
raise ValueError
ValueError

It makes perfect sense because it is looking for a conditioning field in my config, which does not exist because I'm using the generic gan (simple_gan.py) without any conditioning.

[carlpe]

Did you find out how to do it, chinokenochkan?

[enochkan]

@carlpe my apologies for the late reply. I've switched to other source codes (not niftynet) for simple gan.

[carlpe]

@chinokenochkan
May I ask which one you have been most successful?