Question about the classfiers for calculating FID ACC(S) and ACC(C)

[jmliu88]

Hi, thanks for the nice work! I have few questions about the evaluation classifiers while reading the paper. In your paper, you trained “style-aware (S) and content-aware (C) classifiers”. I'm curious about how they have been trained.

Are they trained on both the seen and unseen fonts? If so, what are the training samples? Are all samples from training and testing datasets used to train the classifiers? And what architecture you were using for these classifiers?

Again, LFFont is a great work. Thanks a lot for your time.

James

[8uos]

Hi,

• We used ResNet50 architecture (you can see the code at here).
• We used both seen and unseen fonts for classifier training, but we did not use all the training data. We filtered the training data in three ways;
  1. (Character) We used only the 6,428 characters existing both in GB2312 and our decomposition table for content-aware classifier training.
  2. (Font) We excluded very similar fonts for classifier training to achieve better style accuracy from GT images (> 99%).
  3. (Font) We excluded the fonts which do not include all the 6,428 characters to prevent data skewness.

If you are training your own classifiers, this may help you.

Thanks!

[jmliu88]

Thanks for your reply. It helps!

[SanghyukChun]

Closing the issue, assuming the answer resolves the problem.
Please re-open the issue as necessary.