EC-Net: an Edge-aware Point set.
EC-Net is a point cloud upsampling algorithm based on deep learning, which is collected into our open source algorithm library of point cloud. The original project is run on python-2.7 and tensorflow-1.3, which is in folder './EC-Net-master'. Based on tensorflow-2.x and CUDA-10, we attempt to expande the Tensorlayer version which is in folder './EC-Net-tl'.
Algorithm analysis
please refer to "EC-Net algorithm analysis.doc"
Running commands
-
TF operators
For compiling TF operators, please check tf_xxx_compile.sh under each op subfolder in code/tf_ops folder. Note that you need to update nvcc, python and tensoflow include library if necessary. You also need to remove -D_GLIBCXX_USE_CXX11_ABI=0 flag in g++ command in order to compile correctly if necessary
-
Dijkstra algorithm
Install python-graph library
make install-core
make install-dot
- Train the model
TF: python main.py --phase train --gpu 0 --log_dir ../model/myownmodel
TL: python main.py --phase train --gpu 0 --log_dir ../model/myownmodel_new
- Test and Evaluation
TF: python main.py --phase test --log_dir ../model/pretrain
TL: python main.py --phase test --log_dir ../model/myownmodel_new
Performance comparison
We compare the algorithm performance of different frameworks under tensorflow and tensorlayer respectively. Since the original project did not provide evaluation index calculation code and test data,we adopt test points and mesh from 3PU.
Tensorflow:
pc |
Mean |
std |
min |
max |
a72-seated_jew_aligned |
0.224209 |
0.124589 |
1.34183e-06 |
0.6399 |
A9-vulcan_aligned |
0.177823 |
0.119031 |
6.61831e-07 |
0.577383 |
asklepios_aligned |
0.179195 |
0.121365 |
1.88705e-08 |
0.556526 |
average |
0.193 |
0.121 |
- |
- |
Tensorlayer:
pc |
Mean |
std |
min |
max |
a72-seated_jew_aligned |
0.224586 |
0.125046 |
3.91518e-06 |
0.640771 |
A9-vulcan_aligned |
0.177759 |
0.119361 |
1.80929e-06 |
0.57997 |
asklepios_aligned |
0.179205 |
0.121531 |
5.72039e-06 |
0.561674 |
average |
0.193 |
0.122 |
- |
- |
Contributors
name: Zhang Yongchi
email: zhangych02@pcl.ac.cn