thesouthfrog / stylealign Goto Github PK

Hi! First of all, congratulations on your work!

Could you please provide the runtime of the algorithm for processing one frame (on CPU and GPU)?

Hi, Shengju

Is there any reference link about four autoregressively modeled groups in your repo?

Hello, about the evaluation in your paper, I wonder if you could give me the specific calculation process. I encountered some problems when calculating my result evaluation, thank you.

Hi,

In this repo, kl_loss only calculated the 2-norm between the last layers feature for dec_down and enc_down.
How does it work? Is this implementation equivalent to the formula in paper?

Thank you for your amazing job, do you release your link about AFLW with new 68-point annotation?

Hi!
Your project is great. I have some questions about the dataset section. In your paper, used the AFLW dataset. The original number of the dataset is 19, How to label it as 68 label points? and the WFLW is 98 points. Are 68 points used in actual training? How is it obtained? I am confused about the processing of the data set, I hope you can give me some help. Thank you very much!

Hi! Thanks for your wonderful code.
First, In your pytorch code, the size of input images is 1281283, have you tested for 2562563, can the network provide similar result on different input size?
Second, all face image should be aligned?
Third, I downloaded your wflw-full dataset, how to choose the K sample image for style transform?
Thank you!

Hi, thanks for your work.
Can you provide the pretrained model by torch?
I am very lucky if you reply.

Hi，recently，i have been reading the paper 《Make a Face: Towards Arbitrary High Fidelity Face Manipulation》.
i am still confused about the paper and have a lot of problems. i just wonder whether you have the thought of open source of this paper. i will appreciate it very much if you do.
thankyou very much.

Hi,
Do you plan to release pretrained checkpoints for testing?
Many thanks.

Hi,

Thank you very much for sharing the face landmark model. When I run on my own image, the detection results are not very good. Below is the result of a test image. Did I do something wrong? Thank you very much!

Hello,
I am very happy to share your paper's code regarding stylealign.
Currently, I can not find the checkpoint that you trained.
Could you share the checkpoint file that you used in the paper??
Thank you.

Hi, this is a great work. After study, I have several questions:

1. If I wanna improve my model with your augmentation, can I just download your augmented dataset and do the training? Furthermore, is it ok to combine origin images with stylized ones?
2. The images are heavily stylized, which has a gap between images in real life. Will this perform better than just train with origin images?
3. The augmented images has unclear contour, will this cause inaccurate contour points prediction?

Hello. The work is interesting. Could you share the code of Pytorch version?