Skip to content

Latest commit

 

History

History
 
 

01_classification_modelnet

ModelNet10 Classificaiton

This examples shows how to create grid-octree data from meshes and how to train a simple 3D CNN to classify those shapes:

  1. Run python2 create_data.py. This will download the ModelNet10 dataset, unzip the archive and then create for each mesh an grid-octree structure, which is stored to the disk.
  2. Run th train_mn10_r64.lua.
    This will train the network on the generated grid-octrees for 20 epochs.

You can easily modify the create and train script to increase the resolution.