How can i train human nerf on ZJU-MoCap dataset?

[99sphere]

Hi, thank you for sharing your awesome work.

While reading your paper, I checked the experiment on Novel View Reconstructions on public ZJU MoCap dataset. How can i reproduce it?

Especially, I'm curious about the following things.

1. Your proposed preprocessing is applied to SMPL parameters in ZJU MoCap? or did you use it as is?
2. In your training code(train.py), you load scene NeRF for joint training, but i want to train only Human NeRF. Is it possible?
3. How can i convert ZJU MoCap dataset format to Neuman dataset format?

All of these question is about reproducing Figure 12 in you paper.

Thanks again for sharing your wonderful work!

[litchisea]

Hi, thank you for sharing your awesome work.

While reading your paper, I checked the experiment on Novel View Reconstructions on public ZJU MoCap dataset. How can i reproduce it?

Especially, I'm curious about the following things.

1. Your proposed preprocessing is applied to SMPL parameters in ZJU MoCap? or did you use it as is?
2. In your training code(train.py), you load scene NeRF for joint training, but i want to train only Human NeRF. Is it possible?
3. How can i convert ZJU MoCap dataset format to Neuman dataset format?

All of these question is about reproducing Figure 12 in you paper.

Thanks again for sharing your wonderful work!

I also have the same question, how to train videos on the ZJU MoCap dataset? May I ask if you have solved this problem?