raphaelsulzer / dgnn Goto Github PK

View Code? Open in Web Editor NEW

36.0 4.0 5.0 45.41 MB

[SGP 2021] Scalable Surface Reconstruction with Delaunay-Graph Neural Networks

License: MIT License

Python 98.48% Cython 1.05% C 0.03% Shell 0.44%

sgp surface-reconstruction graph-neural-networks delaunay-triangulation point-cloud 3d-reconstruction deep-learning

dgnn's People

Contributors

Stargazers

Watchers

Forkers

zhl8223 kevincain mingzhaolou liuxinren456852 weexp

dgnn's Issues

Trouble setting up the project

hello,
I have been following the instructions step-by-step for setting up the project but when I reached the 3rd step of the installation
--"python utils/setup_libmesh_convonet.py build_ext --inplace"
it gives me this error : "ValueError: 'libmesh\triangle_hash.pyx' doesn't match any files"
even though the file is there just fine. Is there something that I should have installed before this?
Thank you.

Ext compilation

Hello, thank you for your project.
During ext compilation I had an error:
python utils/setup_libmesh_convonet.py build_ext --inplace
Here is the output

Traceback (most recent call last):
  File "/media/maxim/big-500/dgnn/utils/setup_libmesh_convonet.py", line 88, in <module>
    ext_modules=cythonize(ext_modules),
  File "/media/maxim/big-500/envs/dgnn/lib/python3.9/site-packages/Cython/Build/Dependencies.py", line 965, in cythonize
    module_list, module_metadata = create_extension_list(
  File "/media/maxim/big-500/envs/dgnn/lib/python3.9/site-packages/Cython/Build/Dependencies.py", line 815, in create_extension_list
    for file in nonempty(sorted(extended_iglob(filepattern)), "'%s' doesn't match any files" % filepattern):
  File "/media/maxim/big-500/envs/dgnn/lib/python3.9/site-packages/Cython/Build/Dependencies.py", line 114, in nonempty
    raise ValueError(error_msg)
ValueError: 'libmesh/triangle_hash.pyx' doesn't match any files

To compile it I changed the setup_libmesh_convonet.py file:

triangle_hash_module = Extension(
    'utils.libmesh.triangle_hash',
    sources=[
        'utils/libmesh/triangle_hash.pyx'
    ],
    libraries=['m'],  # Unix-like specific
    include_dirs=[numpy_include_dir]
)

I don't know how it affects the rest of the project.

utils/scan with colmap option

I see the provided /utils/scan binary supports colmap sources as shown below. This suggests the binary may take as input the set of colmap sparse reconstruction outputs cameras.bin, images.bin, and points3D.bin.

Is this correct, and could you suggest a sample invocation?

On a connected point, if we begin with a point cloud, can you confirm that we need to provide a starting mesh to utils/scan via -i mesh.off? That is, /utils/scan reports that it requires 1) an input mesh (as .off), which is a bit confusing since the the utility outputs 2) a Delaunay triangulation via virtual scanning as a prerequisite to the 3) final dgnn mesh.

-s [ --source ] arg (=ply)        Data source and options:
-ply
-npz
-colmap
-omvs
-scan,#points,#cameras,std_noise,%outliers
-tt,#scans
-eth

Questions about custom datasets

Hi,

I'm trying to evaluate some custom pointclouds, and had a couple of questions.

In generate_mesh.py, the generate() function has a TODO to make it compatible with open scenes. If I don't have a ground-truth and simply want to create a mesh from a pointcloud, will this affect me?
In generate_mesh.py, the graph_cut() function mentions about using a custom version of pyGo to avoid conversions to int32. How big of a scene will be affected by int32 overflow?
Also, in the mesh-tools, there a function occ2mesh(). Is this usable if the generate_mesh functions cannot handle large scenes?

Thanks in advance and great research!

Eval on a custom point cloud

I'm attempting to process one of the provided point clouds into a mesh via dgnn as a sanity check before trying my own numpy output. The features seem fine (see log below), however run.py gives an error (also shown below). My steps follow:

Set data path in configs/custom.yaml, e.g.: data: /home/kevin/dgnn/data/sample
Extract features with precompiled feat build provided:
./utils/feat -w /home/kevin/dgnn/data/sample/ -i anchor_0 -s npz
Attempt to create mesh:
python run.py -i -c configs/custom.yaml
The error returned is:

Traceback (most recent call last):
File "/home/kevin/anaconda3/envs/dgnn/lib/python3.9/site-packages/munch/__init__.py", line 103, in __getattr__
return object.__getattribute__(self, k)
AttributeError: 'Munch' object has no attribute 'metrics'

The feature extraction log follows:

-----FEATURE EXTRACTION-----

Working dir set to:
-/home/kevin/dgnn/data/sample/

Read NPZ points...
-from anchor_0
-3110 points read
-with sensor
-with normal

Export points...
-to anchor_0.ply
-3110 points

Delaunay triangulation...
-3110 input points
-3110 output points
-39342 finite facets
-19520(+604) cells
-done in 0s

d parameter of Labatu set to -1

Consider turning on scaling with --sn, if your learning data is not yet scaledto a unit cube!

Start learning ray tracing...
-Trace 1 ray(s) to every point
-Ray tracing done in 0s

Export graph...
-to gt/anchor_0_X.npz
-Exported 20124 nodes and 80496 edges
-in 1s

Export 3DT...
-to gt/anchor_0_3dt.npz
-3110 vertices
-39342 facets
-19520 tetrahedra

Export interface...
-to anchor_0_initial.ply
-Before pruning (points, facets): (3110, 39342)
-After pruning (points, facets): (304, 604)
-done in 0s

-----FEATURE EXTRACTION FINISHED in 1s -----

Unable to reconstruct mesh from a custom dataset

I'm trying to reconstruct a mesh from a custom dataset by using the following command:
$ python run.py -i -c configs/custom.yaml

When has_label attribute is 0 inside custom.yaml file the following exception is raised:

READ CONFIG FROM  configs/custom.yaml
SAVE CONFIG TO  /home/p/src/dgnn/data/models/kf96/config.yaml

######## START INFERENCE OF 1 FILES ########

LOAD MODEL /home/p/src/dgnn/data/models/kf96/model_best.ptm

  0%|                       | 0/1 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "/opt/anaconda3/envs/dgnn/lib/python3.9/site-packages/munch/__init__.py", line 103, in __getattr__    
    return object.__getattribute__(self, k)
AttributeError: 'Munch' object has no attribute 'results'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/opt/anaconda3/envs/dgnn/lib/python3.9/site-packages/munch/__init__.py", line 106, in __getattr__
    return self[k]
KeyError: 'results'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/p/src/dgnn/run.py", line 321, in <module>
    inference(clf)
  File "/home/p/src/dgnn/run.py", line 173, in inference
    prediction = trainer.inference(data, subgraph_sampler, clf)
  File "/home/p/src/dgnn/learning/runModel.py", line 414, in inference
    clf.results.OA_test = 0.0
  File "/opt/anaconda3/envs/dgnn/lib/python3.9/site-packages/munch/__init__.py", line 108, in __getattr__
    raise AttributeError(k)
AttributeError: results

It seems results is never assigned.
Replacing results by results_df triggers another exception.
Usage of results is done when has_label attribute is 0:

    if(clf.inference.has_label):
        clf.inference.metrics = Metrics()
        data_inference.batch_x = data_inference.x
        data_inference.batch_gt = data_inference.y

        self.calcLossAndOA(logits_cell, logits_edge, data_inference, clf, clf.inference.metrics)

    else:
        clf.results.OA_test = 0.0

Note: In my environment, I successfully reconstruct reconbench dataset. Also, in reconbench.yaml file, has_label is set to 1.

Textures / vertex color

In your code I see a few flags for mesh coloring via Open3D, however for some of the ETH data treated in your paper, you briefly note that the textured results are via Let there be color! (WAECHTER M., MOEHRLE N., GOESELE M.). Which approach do your favor for your mesh output?

In my tests I'm using your provided ply2npz.py to compile input, which ignores per vertex color in the source files. Is there another supported route that preserves color data to produce a color mesh -- e.g., a naïve per-face color interpolated from the vertex RGB?

raphaelsulzer / dgnn Goto Github PK

dgnn's People

Contributors

Stargazers

Watchers

Forkers

dgnn's Issues

Trouble setting up the project

Ext compilation

utils/scan with colmap option

Questions about custom datasets

Eval on a custom point cloud

Unable to reconstruct mesh from a custom dataset

Textures / vertex color

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent