Generate my own image?

[nvlabs]

Generate my own image?

[layumi]

Hi @kayakaracakaya
Thanks for your attention on our papers.
You mentioned that you want to transfer the pose of the source to target.
Do you want to preserve the identity of the source image or target image?

If you would like to generate the Fig. 5 or Fig.9 in the paper, you may check

visual_tools/show_smooth.py generates Figure 5 of the paper.

visual_tools/show_smooth_structure.py generates Figure 9 of the paper.

For DG-Net, we focus on the changing appearance of the foreground object, i.e., the cloth of the person.
So it is not optimal for changing the pose.

For changing pose, I recommend several papers as follows. You may refer to them. Hope them could help you.

[1] Wei L, Zhang S, Gao W, et al. Person transfer gan to bridge domain gap for person re-identification CVPR 2018
[2] https://github.com/charliememory/Pose-Guided-Person-Image-Generation