3dgan

Data:

Download the training data from the 3D Shapenet website (/3DShapeNets/volumetric_data/).

Files:

/python:

Code for visulization of objects.

File	Description
visualize.py	visualize object represented as voxels using vtk

python3 visualize.py new_chair.mat -u 0.9 -t 0.1 -i 1 -mc 2

/src:

Code for model training and testing.

File	Description
train.py	3d-GAN model training and testing file
dataIO.py	data input output

To train the model

python3 train.py 0 <path_to_model_checkpoint>

To generate chairs

python3 train.py 1 <path_to_trained_model>

/

File	Description
chair_demo.mat	a mat file of chair object generated from the trained 3dgan model's generator
test.py	Transform .mat file into voxels for visualization input (new_chair.mat)

References:

[1] Tensorflow implementation of 3D Generative Adversarial Network. [Github]

[2] MIT Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling [Github]

[3] Princeton 3D ShapeNets: A Deep Representation for Volumetric Shapes