I encountered extremely poor rendering results when trying to run render_novel_pose.py on a trained Stage 2 DynVideo model.
Specific behavior:
- Stage 1 novel pose rendering: Works normally with average quality
- Stage 2 novel pose rendering: Runs successfully but produces extremely poor quality results (eg. the image at the bottom)
Data preparation:
To run novel pose rendering on Stage 2, I have already generated the inp_map for test_pose using the script scripts/gen_pose_map_our_smpl.py.
Questions:
Have the authors and community encountered similar issues, or is this just my personal problem?
Thank you for your help!

I encountered extremely poor rendering results when trying to run
render_novel_pose.pyon a trained Stage 2 DynVideo model.Specific behavior:
Data preparation:
To run novel pose rendering on Stage 2, I have already generated the inp_map for test_pose using the script
scripts/gen_pose_map_our_smpl.py.Questions:
Have the authors and community encountered similar issues, or is this just my personal problem?
Thank you for your help!