O-CNN: Octree-based Convolutional Neural Networks

By Peng-Shuai Wang, Yang Liu, Yu-Xiao Guo, Chun-Yu Sun and Xin Tong.

Internet Graphics Group, Microsoft Research Asia.

Introduction

This repository contains the implementation of O-CNN introduced in our Siggraph 2017 paper "O-CNN: Octree-based Convolutional Neural Networks for 3D Shape Analysis". The code is released under the MIT license.

Citation

If you use our code or models, please cite our paper.

@article {Wang-2017-OCNN,
    title     = {O-CNN: Octree-based Convolutional Neural Networks for 3D Shape Analysis},
    author    = {Wang, Peng-Shuai and Liu, Yang and Guo, Yu-Xiao and Sun, Chun-Yu and Tong, Xin},
    journal   = {ACM Transactions on Graphics (SIGGRAPH)},
    volume    = {36},
    number    = {4},
    year      = {2017},
}

Installation

O-CNN

O-CNN is built upon the Caffe framework and it supports octree-based convolution, deconvolution, pooling, and unpooling. The code has been tested on the Windows 10 x64 (It can be also built on the Ubuntu 16.04). Its installation is as follows:

• Clone Caffe with revision 6bfc5ca8f7c2a4b7de09dfe7a01cf9d3470d22b3
• The code for O-CNN is contained in the directory caffe. Clone and put it into the Caffe directory.
• Follow the installation instructions of Caffe to build the code to get the executive files caffe.exe, convert_octree_data.exe and feature_pooling.exe etc.

NOTE: Compared with the original code used in the experiments of our paper, the code in this repository is refactored for the readability and maintainability, with the sacrifice of speed (it is about 10% slower, but it is more memory-efficient). If you want to try the original code or do some speed comparisons with our O-CNN, feel free to drop me an email, we can share the original code with you.

NOTE: To build the code on the Ubuntu 16.04, you should first manually install the glog and gflag by using the following command: sudo apt-get install libgflags libgoogle-glog-dev, and then add the following line set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} --std=c++11") in the Line 60 of the CMakeLists.txt file.

Octree input for O-CNN

Our O-CNN takes the octree representation of 3D objects as input. The efficient octree data structure is described in our paper. For convenience, we provide a reference implementation to convert the point cloud with oriented normal to our octree format. Furthermore, we also provide a tool to convert the octree file into ply files, which contains the coordinate of the finest leaf nodes and the corresponding normal signal. Note that when the leaf node is empty, the value of normal signal is (0, 0, 0). The code is contained in the directory octree, along with the Microsoft Visual studio 2015 solution file, which can be built to obtain the executable file octree.exe and octree2PLY.exe.

NOTE: To build the octree, the bounding sphere of the object is needed to be computed. The initial version of our code is built upon the bound sphere library from this link. However, we remove it from our code due to the licence issue. To reproduce the results in our paper, it is highly recommended to download the bound sphere library. For more details, please refer to the comments in the file octree/Octree/main.cpp.

O-CNN in Action

The experiments in our paper can be reproduced as follows.

Data preparation

For achieving better performance, we store all the octree inputs in a leveldb or lmdb database. Here are the details how to generate databases for O-CNN.

• Download and unzip the corresponding 3D model dataset (like the ModelNet40 dataset) into a folder.

• Convert all the models (in OBJ/OFF format) to dense point clouds with normals (in POINTS format). For the definition of POINTS format, please refer to the function void load_pointcloud() defined in the file octree/Octree/main.cpp. Note that some OFF files in the dataset may not be loaded by the tools I provided. It is easy to fix these files. Just open them using any text editor and break the first line after the characters OFF. As detailed in our paper, we build a virtual scanner and shoot rays to calculate the intersection points and oriented normals. The executable files and source code can be downloaded here.

• Run the tool octree.exe to convert point clouds into the octree files.

    Usage: Octree <filelist> [depth] [full_layer] [displacement] [augmentation] [segmentation]
        filelist: a text file of which each line specifies the full path name of a POINTS file
        depth: the maximum depth of the octree tree
        full_layer: which layer of the octree is full. suggested value: 2
        displacement: the offset value for handing extremely thin shapes: suggested value: 0.55
        segmentation: a boolean value indicating whether the output is for the segmentation task.
    Usage: Octree2Ply <filelist> [segmentation]
        filelist: a text file of which each line specifies the full path name of a octree file
        segmentation: a boolean value indicating whether the octree is for the segmentation task
  

• Convert all the octrees into a lmdb or leveldb database by the tool convert_octree_data.exe.

O-CNN for Shape Classification

The instruction to run the shape classification experiment:

• Download the ModelNet40 dataset, and convert it to a lmdb database as described above. Here we provide a lmdb database with 5-depth octrees for convenience.
• Download the O-CNN protocol buffer files, which are contained in the folder caffe/examples/o-cnn.
• Configure the path of the database and run caffe.exe according to the instructions of Caffe. We also provide our pre-trained Caffe model in caffe/examples/o-cnn.

O-CNN for Shape Retrieval

The instruction to run the shape retrieval experiment:

• Download the dataset from SHREC16, and convert it to a lmdb database as described above. Note: the upright direction of the 3D models in the ShapeNet55 is Y axis. When generating octree files, please uncomment line 95 in the file octree/Octree/main.cpp and rebuild the code. Here we provide the lmdb databases with 5-depth octrees for convenience, just download the files prefixed with S55 and un-zip them.

• Follow the same approach as the classification task to train the O-CNN with the O-CNN protocal files S55_5.prototxt and solver_S55_5.prototxt, which are contained in the folder caffe/examples/o-cnn.

• In the retrieval experiment, the orientation pooling is used to achieve better performance, which can be perfromed following the steps below.

  • Generate feature for each object. For example, to generate the feature for the training data, open the file S55_5.prototxt, uncomment line 275~283, set the source in line 27 to the training lmdb, set the batch_size in line 28 to 1, and run the following command.

      caffe.exe test --model=S55_5.prototxt --weights=S55_5.caffemodel --blob_prefix=feature/S55_5_train_ 
      --gpu=0 --save_seperately=false --iterations=[the training object number]
    

  Similarly, the feature for the validation data and testing data can also be generated. Then we can get three binary files, S55_5_train_feature.dat, S55_5_val_feature.dat and S55_5_test_feature.dat, containing the features of the training, validation and testing data respectively.

  • Pool the features of the same object. There are 12 features for each object since each object is rotated 12 times. We use max-pooling to merge these features.

      feature_pooling.exe --feature=feature/S55_5_train_feature.dat --number=12 
      --dbname=feature/S55_5_train_lmdb --data=[the data list file name]
    

  Then we can get the feature of training, validation and testing data after pooling, contained in the lmdb database S55_5_train_lmdb, S55_5_val_lmdb and S55_5_test_lmdb.

  • Fine tune the FC layers of O-CNN, i.e. using the solver_S55_5_finetune.prototxt to re-train the FC layers.

      caffe.exe train --solver=solver_S55_5_finetune.prototxt --weights=S55_5.caffemodel
    

  • Finally, dump the probabilities of each testing objects. Open the file S55_5_finetune.prototxt, uncomment the line 120 ~ 129, set the batch_size in line 27 to 1, change the source in line 26 to feature/S55_5_test_lmdb, and run the following command.

      caffe.exe test --model=S55_5_finetune.prototxt --weights=S55_5_finetune.caffemodel 
      --blob_prefix=feature/S55_test_ --gpu=0 --save_seperately=false --iterations=[...]
    

• Use the matlab script retrieval.m, contained in the folder caffe/examples/o-cnn, to generate the final retrieval result. And evaluated it by the javascript code provided by SHREC16.

O-CNN for Shape Segmentation

The instruction to run the segmentation experiment:

• The original part annotation data is provided as the supplemental material of the work "A Scalable Active Framework for Region Annotation in 3D Shape Collections". As detailed in Section 5.3 of our paper, the point cloud in the original dataset is relatively sparse and the normal information is missing. We convert the sparse point clouds to dense points with normal information and correct part annotation. Here is one converted dataset for your convenience, and the dense point clouds with segmentation labels can be downloaded here.

• Run the octree.exe to convert these point clouds to octree files. Note that you should set the parameter Segmentation to 1 when running the octree.exe. Then you can get the octree files, which also contains the segmentation label.

• Convert the dataset to a lmdb database. Since the segmentation label is contained in each octree file, the object label for each octree file can be set to any desirable value. And the object label is just ignored in the segmentation task.

• Download the protocol buffer files, which are contained in the folder caffe/examples/o-cnn. NOTE: as detailed in our paper, the training parameters are tuned and the pre-trained model from the retrieval task is used when the training dataset is relatively small. More details will be released soon.

• In the testing stage, the output label and probability of each finest leaf node can also be obtained. Specifically, open the file segmentation_5.prototxt, uncomment line 458~485, , set the batch_size in line 31 to 1, and run the following command to dump the result.

    caffe.exe test --model=segmentation_5.prototxt --weights=segmentation_5.caffemodel --gpu=0
    --blob_prefix=feature/segmentation_5_test_ --binary_mode=false --save_seperately=true --iterations=[...]
  

• For CRF refinement, please refer to the code provided here. We will provide the automated tool soon.

Acknowledgments

We thank the authors of ModelNet, ShapeNet and Region annotation dataset for sharing their 3D model datasets with the public.

Contact

Please contact us (Pengshuai Wang [email protected], Yang Liu [email protected] ) if you have any problem about our implementation or request to access all the datasets.