Keras implementation of the PointNet 3D classification network: https://github.com/charlesq34/pointnet
- Includes orthogonal regulatization of features
- Optimized for fast training (1D convolutions and fast 2D max-pooling)
- Achieves similar rates of original paper (~88%-89% accuracy)
Download and unzip the modelnet dataset:
wget https://shapenet.cs.stanford.edu/media/modelnet40_ply_hdf5_2048.zip
unzip modelnet40_ply_hdf5_2048.zip
Tested on Ubuntu 16.04, with Keras 2.2.4 and Tensorflow 1.9.0.
Simply run the training script to proceed. The best model will be saved with respect to the validation loss.
python train.py