How Can I use render novel pose use stage2?

[DuanWei1234]

Hello, thanks to your code, I try to run render_novel_pose.py in stage2. I made some changes to load pose_encoder.pth saved in stage2. Also change the pose_batch and transl_batch like stage1, because the original render result is just part of the body.
However, I find there is little different between stage1 and stage2 novel render results(I just used the data neuman_bike and do the same preprocess steps). So, how can I use checkpoint in stage2 to render novel pose?

[DuanWei1234]

I just changed the script again and made sure that the stage2 checkpoint are loaded correctly，but the results in stage1 and stage2 are so similar，why？

[LarkLeeOnePiece]

Hello, did you use the neuman_bike provided by the author or you gennerated it? When I used the provided neuman_bike data, I met something wrong with "smpl_cano_joint_mat.pth". Then I generated the dataset follow the guideline but I got terrible results.

[DuanWei1234]

Hello, did you use the neuman_bike provided by the author or you gennerated it? When I used the provided neuman_bike data, I met something wrong with "smpl_cano_joint_mat.pth". Then I generated the dataset follow the guideline but I got terrible results.

I just do the preprocessing steps for images by myself

[XiaoQiangSHI]

I trained my own dataset and when I use the weights from stage 1 to run render_novel_pose.py, I get normal results. However, when I switch the weights to those from stage 2 and run render_novel_pose.py, I get all white images as results.

Has anyone encountered the same issue? My email is xiaoqiang.s@outlook.com.

[dongals]

I trained my own dataset and when I use the weights from stage 1 to run render_novel_pose.py, I get normal results. However, when I switch the weights to those from stage 2 and run render_novel_pose.py, I get all white images as results.

Has anyone encountered the same issue? My email is xiaoqiang.s@outlook.com.

I'm encountering the same issue, the white output image. Did you solved this problem?