hzykent / vmnet Goto Github PK
View Code? Open in Web Editor NEWImplementation of ICCV2021(Oral) paper - VMNet: Voxel-Mesh Network for Geodesic-aware 3D Semantic Segmentation
License: MIT License
Implementation of ICCV2021(Oral) paper - VMNet: Voxel-Mesh Network for Geodesic-aware 3D Semantic Segmentation
License: MIT License
Hi~
Thanks for the great work. I find your work use Matterport3D datatset and get pretty good results showed in the paper.
But I don’t find any code related to Matterport3D in this repository. I wonder if you can share the Matterport3D related code and give a brief description of how to reproduce the results on Matterport3D.
Looking forward to your reply and this will help a lot.
Hi there,
Thanks for your amazing work!
I would be glad if you could please answer the following question about the usage of QEM in VMNet.
The publication seems to indicate that for vertex contraction only vertices connected by edges are considered. However, in the original QEM publication, the authors also propose selecting vertex pairs for contraction based on their Euclidean distance. They use a threshold value t
for the Euclidean distance.
Using only vertices connected by edges would imply a threshold of t = 0
. In prepare_data.py, Tridecimator from VCGlib is called. If I understand the call correctly, the optional argument -e
is not passed which specifies the threshold. In the Tridecimator application, the threshold t
then defaults to inf
, meaning all pairs of vertices would be eligible for contraction.
Therefore I would like to know: Is contraction of non-connected vertices possible in VMNet?
Thanks a lot for your time, Benjamin
Hi,
I haven't used ScanNet before, so I'm not familiar with it and need your help. Due to my limited memory space, I want to only download the needed parts of the ScanNet. Since you say:
Our method relies on the .ply as well as the .labels.ply files.
so does it mean that I can download the ScanNet only using this two command?
download-scannet.py -o [directory in which to download] --type _vh_clean_2.ply
download-scannet.py -o [directory in which to download] --type _vh_clean_2.labels.ply
In addition, how much memory space is needed to store the processed data?
Hi, I get some issue about
python run.py --test --exp_name test_split --data_path path/to/processed_data
The output logs are:
use_cuda: True
exp_name: test_split
#parameters 17463870
Traceback (most recent call last):
File "run.py", line 279, in <module>
test(exp_name, test_files)
File "run.py", line 139, in test
model.load_state_dict(checkpoint['model_state_dict'])
File "/home/keroro/Program_Files/miniconda3/envs/tt/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1052, in load_state_dict
self.__class__.__name__, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for VMNet:
Unexpected key(s) in state_dict: "Geo_branch.mid_geo.geo_0.lin_edge.weight", "Geo_branch.mid_geo.geo_0.lin_edge.bias", "Geo_branch.mid_geo.geo_1.lin_edge.weight", "Geo_branch.mid_geo.geo_1.lin_edge.bias", "Geo_branch.cd_5.geo_0.lin_edge.weight", "Geo_branch.cd_5.geo_0.lin_edge.bias", "Geo_branch.de5_geo.geo_0.lin_edge.weight", "Geo_branch.de5_geo.geo_0.lin_edge.bias", "Geo_branch.de5_geo.geo_1.lin_edge.weight", "Geo_branch.de5_geo.geo_1.lin_edge.bias", "Geo_branch.cd_4.geo_0.lin_edge.weight", "Geo_branch.cd_4.geo_0.lin_edge.bias", "Geo_branch.de4_geo.geo_0.lin_edge.weight", "Geo_branch.de4_geo.geo_0.lin_edge.bias", "Geo_branch.de4_geo.geo_1.lin_edge.weight", "Geo_branch.de4_geo.geo_1.lin_edge.bias", "Geo_branch.cd_3.geo_0.lin_edge.weight", "Geo_branch.cd_3.geo_0.lin_edge.bias", "Geo_branch.de3_geo.geo_0.lin_edge.weight", "Geo_branch.de3_geo.geo_0.lin_edge.bias", "Geo_branch.de3_geo.geo_1.lin_edge.weight", "Geo_branch.de3_geo.geo_1.lin_edge.bias", "Geo_branch.cd_2.geo_0.lin_edge.weight", "Geo_branch.cd_2.geo_0.lin_edge.bias", "Geo_branch.de2_geo.geo_0.lin_edge.weight", "Geo_branch.de2_geo.geo_0.lin_edge.bias", "Geo_branch.de2_geo.geo_1.lin_edge.weight", "Geo_branch.de2_geo.geo_1.lin_edge.bias", "Geo_branch.cd_1.geo_0.lin_edge.weight", "Geo_branch.cd_1.geo_0.lin_edge.bias", "Geo_branch.de1_geo.geo_0.lin_edge.weight", "Geo_branch.de1_geo.geo_0.lin_edge.bias", "Geo_branch.de1_geo.geo_1.lin_edge.weight", "Geo_branch.de1_geo.geo_1.lin_edge.bias", "Geo_branch.cd_0.geo_0.lin_edge.weight", "Geo_branch.cd_0.geo_0.lin_edge.bias", "Geo_branch.de0_geo.geo_0.lin_edge.weight", "Geo_branch.de0_geo.geo_0.lin_edge.bias", "Geo_branch.de0_geo.geo_1.lin_edge.weight", "Geo_branch.de0_geo.geo_1.lin_edge.bias".
What are the Unexpected key(s)
in state_dict ? does the VMNet not defined?
vcglib/apps/tridecimator
and vcglib/apps/sample/trimesh_clustering
, I add environment path by:export PATH=$PATH:/path/to/vcglib/apps/tridecimator:/path/to/vcglib/apps/sample/trimesh_clustering
# create links
sudo ln -s /path/to/vcglib/apps/tridecimator/tridecimator /usr/local/bin
sudo ln -s /path/to/vcglib/apps/sample/trimesh_clustering/trimesh_clustering /usr/local/bin
But run the preprocess, there is core dumped:
in_path:../scannet/VMtest
out_path:../scannet/VMNet_data/train/
[0.02, 0.04, 30, 30, 30, 30, 30]
Processing ../scannet/VMtest/scene0000_00/scene0000_00_vh_clean_2.ply
curr_dir: ../scannet/VMNet_data/train/scene0000_00
trimesh_clustering: ../../../vcg/simplex/vertex/component.h:75: vcg::vertex::EmptyCore<TT>::ColorType& vcg::vertex::EmptyCore<TT>::C() [with TT = MyUsedTypes; vcg::vertex::EmptyCore<TT>::ColorType = vcg::Color4<unsigned char>]: Assertion `0' failed.
Aborted (core dumped)
trimesh_clustering: ../../../vcg/simplex/vertex/component.h:75: vcg::vertex::EmptyCore<TT>::ColorType& vcg::vertex::EmptyCore<TT>::C() [with TT = MyUsedTypes; vcg::vertex::EmptyCore<TT>::ColorType = vcg::Color4<unsigned char>]: Assertion `0' failed.
Aborted (core dumped)
multiprocessing.pool.RemoteTraceback:
"""
Traceback (most recent call last):
File "/home/keroro/Program_Files/miniconda3/envs/tt/lib/python3.7/multiprocessing/pool.py", line 121, in worker
result = (True, func(*args, **kwds))
File "/home/keroro/Program_Files/miniconda3/envs/tt/lib/python3.7/multiprocessing/pool.py", line 44, in mapstar
return list(map(*args))
File "prepare_data.py", line 226, in process_frame
old_vertices=vertices[-1])
File "prepare_data.py", line 173, in quadric_error_metric
'.ply', '.csv'), old_vertices=old_vertices, new_vertices=vertices_l)
File "prepare_data.py", line 78, in csv2npy
with open(in_file_path, 'r') as csvfile:
FileNotFoundError: [Errno 2] No such file or directory: '../scannet/VMNet_data/train/scene0000_00/curr_mesh.csv'
"""
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "prepare_data.py", line 295, in <module>
pf_pool.map(process_frame_p, file_paths)
File "/home/keroro/Program_Files/miniconda3/envs/tt/lib/python3.7/multiprocessing/pool.py", line 268, in map
return self._map_async(func, iterable, mapstar, chunksize).get()
File "/home/keroro/Program_Files/miniconda3/envs/tt/lib/python3.7/multiprocessing/pool.py", line 657, in get
raise self._value
FileNotFoundError: [Errno 2] No such file or directory: '../scannet/VMNet_data/train/scene0000_00/curr_mesh.csv'
Where should I set the correct environment path for vcglib/apps/tridecimator
and vcglib/apps/sample/trimesh_clustering
?
Line 47 in 14e2c25
Hi, authors, thanks for sharing your work.
When I tried to train VMNet with my own training data, the valid_idxs is not meet the assert condition, like
Line 158 in 6181683
I think the problem is related to my own data, but the specific reason for such an issue is unclear. Do you have any suggestions about that?
PS: I use the prepare_data.py script to preprocess my own data, as same as the preprocessing procedure for the ScanNet dataset.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.