search

bennyguo / instant-nsr-pl

Neural Surface reconstruction based on Instant-NGP. Efficient and customizable boilerplate for your research projects. Train NeuS in 10min!
MIT License
856 stars 84 forks source link

Want to be linked to nerfacc repo? #1

Closed liruilong940607 closed 2 years ago

liruilong940607 commented 2 years ago

Awesome job Yuanchen! I was thinking of doing an example in nerfacc with NeuS. But you already got it worked even with pytorch-lightning!!

Do you want to be linked to nerfacc repo for showcase?

BTW I noticed you are relying on alpha marching & rendering, which is currently from your fork of nerfacc. I just add this support in nerfacc>=0.2.2, along with relaxing the pytorch requirement. Check it out if interested! https://github.com/KAIR-BAIR/nerfacc/pull/94

bennyguo commented 2 years ago

Thanks Ruilong! That would be great!

I already update the dependency to your official nerfacc repo in the latest commit, cheers 🎉

liruilong940607 commented 2 years ago

Linked! Great job again!

Jiakui commented 2 years ago

Hi Ruilong,

It looks for me that, although much faster, the neus with hash encoding is much worse than the vallina MLP version neus (I tried both instant-nsr and instant-nsr-pl , they are both not as good as neus, so I guess hash encoding should be responsible for the degradation of the reconstruction quality ) . Do you have any idea why ?

Thanks!