How to train NVIDIA-long-release dataset?

[sjdfkjawe]

Hello, thank you so much for making your work open source. May I ask whether it is possible to train NVIDIA-long-release., such as balloon1 dataset, directly in the way of DynibarFF model, I notice you didn't open the train.py file for NVIDIA-long-release. Or if I can train the NVIDIA-long-release dataset in the DynibarMono way after preparing pre-processed files such as disp，masks ， flow same as kid-running dataset.

Looking forward to your reply very much，Thank you.

[zhengqili]

Hi, The major difference is for nvidia dataset, you don't need virtual source view, so you can remove this part in the data loader and rendering function.

[CTouch]

Thanks for your great work!

I have the similar question about train/render the data in nvidia_long_realease. I don't really understant that 'nvidia dataset doesn't need virtual source view,'. If so, where can we set the novel view poses?

Thanks!