Open serizawa-04013958 opened 4 months ago
Hi! I'll double-check to confirm the reason for this. The results from Stage 2 aren't significantly better than those from Stage 1. For Stage 2 rendering, additional posemap input is required.
Thank you so much!!
I got it. Also in current code, stage2 only save pose_encoder.pth, not net.pth.
Is it correct? because when we render, render_novel_pose.py always ask us to load net.pth even stage 2.
Hi! Yes, you're right. The render_novel_pose.py is only for stage one.
OK, I wait for your update. Thank you so much
I faced same problems to the other issues. How can i render with stage2? my stage1 results with neuman dataset was not so good.
hi, are you download the neuman dataset and process it yourself ? because i find only neuman-bike are provided. looking for your reply , thank you
yes. I followed instantavatar preprocessing.
yes. I followed instantavatar preprocessing.
hi,i wonder how to obtain the additional pose map, when in stage2 the train.py ask me to get the /inp_map/inp_posemap_128_00000052.npz
looking for your reply , thank you
I faced same problems to the other issues. How can i render with stage2? my stage1 results with neuman dataset was not so good.![00212](https://github.com/aipixel/GaussianAvatar/assets/78073281/d4224d70-9824-4609-98dc-2de623a6f6e8)