DealWithSHREC

2018/07/31 Jiadai Sun sunjiadai@foxmail.com

The main purpose of this project is to generate a data set in HDF5 format for PointNet as input.

The directory structure looks messy. I will continue to improve it.

Need Dimension Normalization	Contrast Query and Database	Sampling from Query

run python copymodel.py copy 3308 shrec17_data_jan27/shapenet/models/***.obj to model_obj/***.obj
run obj2pcd.sh , use mesh_sampling_self to render model_obj/***.obj to PointCloud,Generate a set number of point clouds and save to model_ply/***.pcd, use pcl_pcd2ply transform model_ply/***.pcd to model_ply/***.ply
run ply_normalize.sh use normalization to normalizemodel_ply/***.ply including location_normalization and dimension_normalization , save the model_ply/***.ply after normalizing to model_ply_normalize/***.ply
run python wrire_hdf5.py , save PointCloud and Label to hdf5 file, prepare for the input of A
use some scripts to help me quickly process large amounts of data
model.txt(Part of train.csv) contain 3308 models from shapenet, named wss.***

pointdeal_cpp/***.cpp
- normalization.cpp
  
  Dimension Normalization and Location Normalization, without Pose Normalization
- resampling.cpp
  
  if the number of PointClode is less than 2048, we should upsampling and downsampling to 2048
- mesh_sampling.cpp
  
  Used for desampling the number of Point Cloud to -n_samples 2048
  
  If I use the pcl_mesh_sampling in script to deal many *.ply / *.pcd / file, need to close the windows and go on to the next one. I re-compile the file and can use the parameter -no_vis_result to Prohibit window pop-up
write_hdf5.py

Read the csv file and PointCloud generate by PCL
get_hdf5_data.py

This model is used to see some information and check the correctness of *.h5
copymodel.py & copyply.py

copy the shapenet model in ObjectNN and *.ply in scenenn we will use to this folder

Unused File

MaxChanger/SHREC-Process