What is your operating environment and the version of the software?

[Outhbreaker]

I want to reproduce your experiment. But there have been some problems in the debugging process. There are even code errors. I think this will definitely not be your problem. So I want to ask the specific environment running version (Ps. My environment is configured according to Readme)
Such as: tensor_gt[batch,0,:,:] = tensor_all.data[batch,self.out_id[batch],:,:]
IndexError: list index out of range

[azadis]

I think the issue is due to a wrong batch size specified here https://github.com/azadis/MC-GAN/blob/master/scripts/train_StackGAN.sh#L37
This batch size should be equal to the number of letters of your input $DATA. For instance, for a given word FROZEN with its own font style, BATCHSIZE should be set to 6.

[Outhbreaker]

I am so happy that I could receive your reply so soon.

Then I want to talk about my current understanding. It may be that I have a problem with the understanding of the project or what misunderstanding the configuration of the environment, which leads to various states.

Before I used the old version of the pytorch，the problem I will put the picture in the attachment.And also I send the new problem to you.

This is my current environment configuration (Ubuntu 16.04 + torch 1.0 + torchvision 0.21 others using pip install)

After modifying the environment version, I probably changed 3 bugs to make train_cGAN.sh run successfully,so I think it should be caused by some deviations from my version. I want to know the specific environment configuration on your side.

Today, I follow your introduction to open train_StackGAN.sh and modify the batchsize=5. And I execute the statement "./scripts/train_StackGAN.sh BLOOD", but the problem is still here.

After that，I also want to know README.md: ./scripts/train_StackGAN.sh ${DATA} this ${DATA} should be replaced by the specific folder name such as BLOOD If do that, that means lots of folders need run one by one.

In short, I am very grateful to hear from you and I will look forward to your next reply.

[azadis]

Sorry for the confusion. For BLOOD, I think you should set your batch size to 4. This will be the number of unique given letters.
I don't think there is an issue with your environment and torch or torchvision versions.

The answer to your second question is actually yes. You need to train the end-to-end model for transferring the style of each font separately. This will be done through the train_StackGAN.sh file with the name of your data folder. Prior to that, you should train the Glyph Network on all fonts for learning the glyph shapes by running train_cGAN.sh.

[Outhbreaker]

Thanks for your answer. I have finished running train_cGAN.sh. And I also take the photo of running test_cGAN.sh. I follow your suggestion to change the batch size, but the problem is still there. Sorry, I don't want to bother you. I really don't know how to solve it. By the way, such as ft2_2, is this folder batch size is the same as the BlOOD? I will look forward to your reply.

[azadis]

I think your issue is with the running path or data path. You are probably either running the code from a wrong path (do you run bash ./scripts/train_StackGAN.sh BLOOD ?) or your in-the-wild fonts are not downloaded to the correct path. Have you run ./datasets/download_font_dataset.sh public_web_fonts ? BLOOD directory should exist under ../datasets/public_web_font.

[Outhbreaker]

I downloaded the date except the optional Others have already downloaded in my repertory.
I show you the data picture.


Also, I follow what you said (In fact as before I did it the same as what you said )but the result is still as usual.

[liukboy]

@Outhbreaker I meet the same issue with you,and now how do you settled it?Thank you.

[Outhbreaker]

@liukboy I don't know how to do it. I still wait the reply from the author. By the way, are you Chinese college students？

[azadis]

The datasets directory containing the font files should not be under MC-GAN! Your current font files path is not correct. You should be able to do the following:
cd path_to_MC-GAN
ls ../datasets/public_web_fonts