PointNet AutoEncoder using chainer

Introduce

AutoEncoder using PointNet and chainer. PointNet-AutoEnocoder(TensorFlow) which PointNet author created is here.

Installation

Please install Chainer (and cupy if you want to use GPU) beforehand.
Furthermore, Operation check is described on comments.

# chainer version 5.0.0
pip install chainer
# cupy-cuda92 version 5.0.0
pip install cupy-cuda92

Also, some extension library is used in some of the code,

# Chainer Chemistry version 0.4.0
git clone https://github.com/pfnet-research/chainer-chemistry.git
pip install -e chainer-chemistry
# ChainerEX version 0.0.1
git clone https://github.com/corochann/chainerex.git
pip install -e chainerex

Download Data

ShapeNetPart dataset is available HERE. Simply download the zip file and move the shapenetcore_partanno_segmentation_benchmark_v0 folder to data.

Besides this, you can use command to download ShapeNetPart dataset.

python dataset.py -d true

Train

You can simply execute train code with GPU.

python train.py -g 0

Check using trained model and viewer

You can view a ground truth data and then view an output data. In order to view next data, Please press q.

python test.py

References

1. Charles R. Qi, Hao Su, Kaichun Mo, Leonidas J. Guibas. PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation. CVPR 2017.
2. "charlesq34/pointnet-autoencoder: Autoencoder for Point Clouds". Github. https://github.com/charlesq34/pointnet-autoencoder, (accessed 2018-12-5).
3. "corochann_chainer-pointnet_ Chainer implementation of PointNet, PointNet++, KD-Network and 3DContextNework". Github. https://github.com/corochann/chainer-pointnet, (accessed 2018-11-13).