• EC-Net: an Edge-aware Point set.
• • Algorithm analysis
  • Running commands
  • Performance comparison
  • Contributors

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

1. 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

2. Dijkstra algorithm
   Install python-graph library

make install-core
make install-dot

3. 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

4. Test and Evaluation

TF: python main.py --phase test --log_dir ../model/myownmodel_new
TF: python main.py --phase test --log_dir ../model/pretrain

Performance comparison

We compare the algorithm performance of different frameworks under tensorflow and tensorlayer respectively. The main evaluation metric is P2F. Note: Since the original project did not provide evaluation index calculation code and test data，we adopt P2F metric on 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

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

Contributors

name: Zhang Yongchi
email: zhangych02@pcl.ac.cn