sxyu / volrend Goto Github PK

Required openGL version: 4.3
But Apple has deprecated openGL and latest supported openGL version is 4.1.

Is there any way to run with openGL 4.1?

Hi, thank you for work.

I wanted to see if I can produce .png images. I downloaded pretrained tree, intrinsics and poses. Then I ran

./volrend_headless /home/hyo26751/Downloads/tree.npz -i /home/hyo26751/Downloads/drums/intrinsics.txt /home/hyo26751/Downloads/drums/pose/* -o ../
INFO: Use NeRF camera convention
INFO: Data format SH16
INFO: Scale 0.445217 0.588506 0.68048943.2574615479 ms per frame
23.1173992157 fps

which produced a set of blank images. Possible mistakes I have made?

hi sxyu,
When running the code on ubuntu20.04, I encountered a cuda-related problem. I tried the 11.0 and 11.1 versions of cuda and reported the error in the figure below. But It can run normally with the same configuration on the computer of my classmate.

When I try to run with window11, but it automatically runs on integrated graphics card while the computer has a discrete graphics card.

I would be grateful if you could help me out with these two issues.
@sxyu

I trained nerf-sh model with chair dataset. After extraction and optimization, I got the tree_opt.npz. However, I loaded this npz with webgl viewer, I got an incorrect viewing result as follows:

If I launch this npz with cuda-based volrend viewer, I get correct viewing result.
Furthermore, I downloaded the uncompressed NeRF-synthetic files used for the web demo, the npz file is ok when I load with webgl viewer. So, I have a question. Is there any difference between the tree_opt.npz and you provided uncompressed npz model ?
@liruilong940607 @sxyu

I want to make Plenoctree streaming server because plenoctree web viewer is pretty laggy and lack of mobile support.

So i tried to make headless version w/o CUDA, but it wont works w/o GLFW window. is there any solution?

I use the script/compress_octree.py to compress the full resolution tree files, and use it in the web demo I build with emcmake.
But it get stuck

however, when I use the The uncompressed NeRF-synthetic files you provided, it succeed.

How can I compress the npz by myself in order to use this npz in web demo?

Which leads to inconsistencies when dealing with RGBA data. Happy to submit a PR either adding sigmoid to volrend or removing it from svox if you have a preference either way.

In ubuntu 22.04, gcc-10, cuda-11.3, compile errors when run 'make -j12' command. I have installed the libc6-dev, and locate the 'libc6-dev:amd64: /usr/lib/x86_64-linux-gnu/librt.a'.

i want to running volrend in web alocated in the cloud. how can set private ip host and port?

What is the difference between plenoctree and octree？

Hi, thanks for your work!
When I run the code, the error occurred.

what should I do? pls

Hello, thank you for you and your team's work. I see the other repositories connected to the Plenoctree NeRF paper are under the BSD-2-Clause license and was wondering if there was a preferred licensing for volrend as well. Thank you!

hey,any software to achieve this?

As title descript, when i compress the .npz with bits 8, and open the npz with volrend, the program crash

$ ./volrend tree.npz

INFO: Data format SH16
INFO: Decoding quantized colors
[1]    1060447 segmentation fault (core dumped)  ./volrend

The volrend.exe file is under build/Release after build, so there will be error if you run ./volrend directly when your current path is just build. When your current command path is under build, the correct way is cd Release first and then run ./volrend, or you can just run ./Release/volrend instead.
Besides, if you have the .npz files, you can just put it under Release file, and run ./volrend name.npz.

void dda_unit(vec3 cen, vec3 invdir, out float tmax) {
    float t1, t2;
    tmax = 1e9f;
    for (int i = 0; i < 3; ++i) {
        t1 = - cen[i] * invdir[i];
        t2 = t1 +  invdir[i];
        tmax = min(tmax, max(t1, t2));
    }
}

I may have trouble understanding it...could you help me..many thanks!..
It looks like the t1 is trying to go back to the origin...

Hi, I have error with cuda when run ./volrend

Is it due to insufficient RAM ?

Hi, when trying to display my custom made plenoctree in the web viewer I get the following error:

It works in the desktop based volrend version, anyone got a solution?

I use pip install svox to install svox. However, when I import svox in python, it raises the following warning:

UserWarning: CUDA extension svox.csrc could not be loaded! Operations will be slow.
Please do not import svox in the SVOX source directory.
warn("CUDA extension svox.csrc could not be loaded! "

Platform: ubuntu~16.04.10
Cuda Version: 10.1

Thanks!

GLX: GLX version 1.3 is required

glxinfo|grep version
server glx version string: 1.2
client glx version string: 1.4
GLX version: 1.2
OpenGL version string: 1.1.0

I tried to find the reason, but it failed in the end.Do you understand why this is?Thanks!!!!!!!

Hi, Thanks for this awesome project. I wanted to ask if it is possible to view an already trained NeRFin Nerfstudio via PlenOctrees viewer in a non-GPU machine?
Thanks again