https://github.com/dakup/handposeshapevae
Experimental hub; various VAE implementations for unsupervised disentanglement of hand pose and shape latent space
https://github.com/dakup/handposeshapevae
Last synced: about 1 year ago
JSON representation
Experimental hub; various VAE implementations for unsupervised disentanglement of hand pose and shape latent space
- Host: GitHub
- URL: https://github.com/dakup/handposeshapevae
- Owner: DaKup
- Created: 2020-08-02T14:47:10.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2020-08-09T16:54:15.000Z (almost 6 years ago)
- Last Synced: 2025-02-01T11:43:56.749Z (over 1 year ago)
- Language: Python
- Size: 71.1 MB
- Stars: 4
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# HandPoseShapeVAE
(not actively maintained)
Experimental framework for unsupervised hand pose and shape disentanglement
Thanks to
- https://github.com/moberweger/deep-prior-pp
- https://github.com/rtqichen/beta-tcvae
- https://github.com/Knight13/beta-VAE-disentanglement
- https://github.com/AntixK/PyTorch-VAE
## Getting Started
These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. See deployment for notes on how to deploy the project on a live system.
### Prerequisites
What things you need to install the software and how to install them
- [PyTorch 1.1](https://pytorch.org/)
- [CUDA 10](https://developer.nvidia.com/cuda-downloads)
- [TensorboardX](https://github.com/lanpa/tensorboardX)
- [2015 MSRA Hand Gesture Dataset](https://jimmysuen.github.io/)
### Setup
(In the following, the working directory should be the same location as this readme file)
Create the following directories:
```txt
../data/preprocessed
../models
```
Download/Save dataset as
```txt
../data/MSRA15/P0
...
../data/MSRA15/P8
```
## Preprocess
Import sequences into rendered format
```txt
preprocess/import_msra.py --input-dataset "../data/MSRA15" --label-name "msra"
```
Merge rendered sequences
```txt
preprocess/merge_rendered.py
```
Split dataset into train, validation and test datasets
```txt
preprocess/preprocess_rendered.py --input-dataset "../data/msra_merged.npz" --label-name "msra_merged" --random
```
## Train
```txt
main.py --basename "HandPoseShapeVAE_MSRA" --model "bvae" --auto-load --auto-save --beta 15.0 --train --train-dir "../data/preprocessed/train_msra_merged.npz" --validation-dir "../data/preprocessed/validate_msra_merged.npz" --test-dir "../data/preprocessed/test_msra_merged.npz"
```
## Test
```txt
main.py --basename "HandPoseShapeVAE_MSRA" --model "bvae" --auto-load --auto-save --beta 15.0 --test --train-dir "../data/preprocessed/train_msra_merged.npz" --validation-dir "../data/preprocessed/validate_msra_merged.npz" --test-dir "../data/preprocessed/test_msra_merged.npz"
```
## Experiments
```txt
main.py --basename "HandPoseShapeVAE_MSRA" --model "bvae" --auto-load --auto-save --beta 15.0 --experiments --traverse-latents --train-dir "../data/preprocessed/train_msra_merged.npz" --validation-dir "../data/preprocessed/validate_msra_merged.npz" --test-dir "../data/preprocessed/test_msra_merged.npz"
```
```txt
main.py --basename "HandPoseShapeVAE_MSRA" --model "bvae" --auto-load --auto-save --beta 15.0 --experiments --train-joints --train-dir "../data/preprocessed/train_msra_merged.npz" --validation-dir "../data/preprocessed/validate_msra_merged.npz" --test-dir "../data/preprocessed/test_msra_merged.npz"
```
