Causal_Disentangle

How to generate synthetic 3D shape dataset

Introduction

There are 6 attributes in total, each has several possible values.

'floor_hue', 'wall_hue', 'object_hue', 'scale', 'shape', 'orientation'

Here we use the object_hue and shape as factors of interests.

There are mainly three files related to the generative process.

Data Sampler

Causal_Disentangle/3dshape_dataset/dataset_3d_shape.py
/Causal_Disentangle/prepare_3d_shape.ipynb
/Causal_Disentangle/datasets.py

Use the jupyternotebook as a reference.

We first create a folder to store the generated images. Then we use a csv file to store the attributes and labels. labels are structured as ["label_name", actual_label(integer)]

Name		Name	Last commit message	Last commit date
Latest commit History 48 Commits
.ipynb_checkpoints		.ipynb_checkpoints
3dshape_dataset		3dshape_dataset
CANDLE_dataset		CANDLE_dataset
data		data
models		models
outputs		outputs
.gitignore		.gitignore
CANDLE_label.txt		CANDLE_label.txt
README.md		README.md
Untitled.ipynb		Untitled.ipynb
causalvae.py		causalvae.py
celebA_causalvae_exp.py		celebA_causalvae_exp.py
celebA_exp.py		celebA_exp.py
datasets.py		datasets.py
doVAE.ipynb		doVAE.ipynb
example.ipynb		example.ipynb
metrics.py		metrics.py
multigpu_example.py		multigpu_example.py
ood.py		ood.py
plot_images.ipynb		plot_images.ipynb
prepare_3d_shape.ipynb		prepare_3d_shape.ipynb
run_VAE.py		run_VAE.py
run_causal_vae.sh		run_causal_vae.sh
run_doVAE.py		run_doVAE.py
run_vae_exp.sh		run_vae_exp.sh
sqrtm.py		sqrtm.py
vae.py		vae.py

xliu1231/Causal_Disentangle

Folders and files

Latest commit

History

Repository files navigation

Causal_Disentangle

How to generate synthetic 3D shape dataset

Introduction

About

Resources

Stars

Watchers

Forks

Languages