How to test with your own images?

[molo32]

How to test with your own images?

I have jpg images of faces, how inpainting?

[tiwarikaran]

I have the same issue lol

[zsyzzsoft]

I will add support for this in the next week. But note that the training images of the FFHQ dataset are preprocessed and aligned; custom images will hardly meet this requirement unless you do the same preprocessing as theirs. The Places2 model would be more generalizable and can be used to inpaint your own photos.

[tiwarikaran]

No no, I am very much willing to sit down and do the pre-processing.

[tiwarikaran]

File "/content/co-mod-gan/dnnlib/submission/submit.py", line 338, in submit_run raise RuntimeError("Invalid task name. Probable reason: unacceptable characters in your submit_config.run_desc. Task name must be accepted by the following regex: " + docker_valid_name_regex + ", got " + submit_config.task_name)

RuntimeError: Invalid task name.

• Probable reason:
  • unacceptable characters in your submit_config.rundesc. Task name must be accepted by the following regex: ^[a-zA-Z0-9][a-zA-Z0-9.-]+$, got root-00002-co-mod-gan-/content/drive/MyDrive/data_ajanta-1gpu

[zsyzzsoft]

File "/content/co-mod-gan/dnnlib/submission/submit.py", line 338, in submit_run raise RuntimeError("Invalid task name. Probable reason: unacceptable characters in your submit_config.run_desc. Task name must be accepted by the following regex: " + docker_valid_name_regex + ", got " + submit_config.task_name)

RuntimeError: Invalid task name.

• Probable reason:

•   unacceptable characters in your submit_config.run_desc.  Task name must be accepted by the following regex: ^[a-zA-Z0-9][a-zA-Z0-9_.-]+$, got root-00002-co-mod-gan-/content/drive/MyDrive/data_ajanta-1gpu

In your case, you should pass `--data-dir=/content/drive/MyDrive --dataset=data_ajanta'

[zsyzzsoft]

Now one can create a custom dataset with dataset_tools/create_from_images and then run demo with it!