xth430 / deforming-nerf Goto Github PK

I read the code of the paper.

In svox2.py, I don't understand grid2world function.
roffset = self.radius * (1.0 / gsz - 1.0) + self.center
Why do I need to add self.radius * (1.0 / gsz)?

Hi, I moved some verteces of your provided deformed chair cage ,and also keep the vertex order when export.
And then I ran:

python render_imgs_deform.py ckpt/nerf_lego/ckpt.npz ../data/nerf_synthetic/lego/ -c configs/syn.json --interpolate --cam_id 64

I got

Traceback (most recent call last):
  File "render_imgs_deform.py", line 93, in <module>
    cage_source, cage_target = load_cages(args)
  File "/root/deforming-nerf/opt/deformation/util.py", line 19, in load_cages
    assert cage_target.is_watertight, "Cage (target) is not watertight!"
AssertionError: Cage (target) is not watertight!

It seems the cage I deformed became not wartertight anymore,but I look it up in blender, it seems nowhere is unclosed.
Could you give me some advices help to move the vertices while keep the mesh wartertight?

Hi!Thanks for your great work!
I captured about 90 pictures(360°） around an object like this(removed background manualy in photoshop):

and get the c2w metrics from directly from colmap
render a static scene using syn.config,but the results are very bad:


Could you please give me some advices about where could go wrong

And I tried dtu dataset scan83,the output fine.obg is very odd like this:

It is nothing like the scene and the size of ckpt.npz of this scene is about 1.3GB, more less than your provided ckpt.npz of scan83,could you tell me how to get the resonable output?

Hello. Thank you for your wonderful work.

I have a question about training the custom dataset.
I have images, mesh from those images, and deformed mesh, and I want to render them with your deforming-nerf code.
The problem is my meshes are not cage-mesh. How can I transform regular mesh into cage mesh?

Could you please give some instructions regarding the custom dataset training part including how to build cage-mesh?

Thank you

Hello, thanks for your excellent work!
But when I follow the MANUAL.md and try to estimate the cage for the lego case, I get the following results:

But in the given ckpt file, the cage for lego is:

Is there any way to improve the estimated cage for custom images?

When I try to extract mesh from nerf-lego dataset, I ran

python cage/extract_mesh.py ckpt/nerf_lego/ckpt.npz

and then I got

[INFO] loading Plenoxel model from: ckpt/nerf_lego/ckpt.npz
Traceback (most recent call last):
  File "cage/extract_mesh.py", line 18, in <module>
    grid = svox2.SparseGrid.load(args.ckpt, device=device)
  File "/root/miniconda3/envs/deforming-nerf/lib/python3.8/site-packages/svox2/svox2.py", line 1560, in load
    z = np.load(path)
  File "/root/miniconda3/envs/deforming-nerf/lib/python3.8/site-packages/numpy/lib/npyio.py", line 420, in load
    ret = NpzFile(fid, own_fid=own_fid, allow_pickle=allow_pickle,
  File "/root/miniconda3/envs/deforming-nerf/lib/python3.8/site-packages/numpy/lib/npyio.py", line 187, in __init__
    _zip = zipfile_factory(fid)
  File "/root/miniconda3/envs/deforming-nerf/lib/python3.8/site-packages/numpy/lib/npyio.py", line 103, in zipfile_factory
    return zipfile.ZipFile(file, *args, **kwargs)
  File "/root/miniconda3/envs/deforming-nerf/lib/python3.8/zipfile.py", line 1269, in __init__
    self._RealGetContents()
  File "/root/miniconda3/envs/deforming-nerf/lib/python3.8/zipfile.py", line 1336, in _RealGetContents
    raise BadZipFile("File is not a zip file")
zipfile.BadZipFile: File is not a zip file

The ckpt.npz file is directly downlowded from the provided link and using tar -zxvf to unzip it.
I 'm using ubuntu20.04,pytorch 1.11.0,cuda 11.3

Congratulations！I wanna ask when will your team release the code. Appreciated!