Error reported when using pre trained model!

[VerseWei]

hello,

Thank you for this great work. I encountered some errors while using the pre trained model. Can you help me?

For dtu Scan24, I will ckpt 300000. pth is placed in./exp/dtu Scan24/neuda Under womask/checkpoints, add image, mask, and cameras Spheres. npz is placed in data/DTU/dtu Scan24, and then we encountered this problem.

(neuda) wxs@cvrs:~/3D_Recon/NeuDA$ python extract_mesh.py --case dtu_scan24 --conf ./conf/neuda_womask.conf --eval_metric /home/wxs/anaconda3/envs/neuda/lib/python3.9/site-packages/torch/functional.py:504: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1682343997789/work/aten/src/ATen/native/TensorShape.cpp:3483.) return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined] [INFO] SDF Network dims: [51, 256, 256, 256, 260] [extract_mesh.py:68 - extract_mesh() ] Find checkpoint: ckpt_300000.pth Load data: Begin [-1.01 -1.01 -1.01] [1.01 1.01 1.01] Load data: End dilation... Mask dilation: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 49/49 [00:05<00:00, 9.28it/s] Low resolution evaluation: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10/10 [00:00<00:00, 93.25it/s] Rasterization: 0%| | 0/49 [00:00<?, ?it/s] Traceback (most recent call last): File "/home/wxs/3D_Recon/NeuDA/extract_mesh.py", line 154, in extract_mesh(args) File "/home/wxs/3D_Recon/NeuDA/extract_mesh.py", line 104, in extract_mesh mesh = get_mesh.get_surface_high_res_mesh( File "/home/wxs/3D_Recon/NeuDA/evaluation/get_mesh.py", line 99, in get_surface_high_res_mesh face_mask = face_mask & mesh_filtering.visible_mask(mesh_low_res, cams, nb_visible=1).cpu().numpy() File "/home/wxs/3D_Recon/NeuDA/evaluation/mesh_filtering.py", line 43, in visible_mask pix_to_face, zbuf, bar, pixd = meshRasterizer(meshes) TypeError: cannot unpack non-iterable Fragments object

[bxcharlie]

Hello, I have the same problem, how did you solve it?

[BwCai]

Sorry for the late reply. This may be incompatible between different pytorch3d versions. Fragments (returned by meshRasterizer) have changed from NamedTuple to dataclass. If you use recent pytorch3d version (>=0.7.0), you need to change the mesh_filtering.py line 43 as follows:

mesh_frag = meshRasterizer(meshes)
pix_to_face, zbuf, bar, pixd = mesh_frag.pix_to_face, mesh_frag.zbuf, mesh_frag.bary_coords, mesh_frag.dists