Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/amirbar/speech2gesture
code for training the models from the paper "Learning Individual Styles of Conversational Gestures"
https://github.com/amirbar/speech2gesture
Last synced: 6 days ago
JSON representation
code for training the models from the paper "Learning Individual Styles of Conversational Gestures"
- Host: GitHub
- URL: https://github.com/amirbar/speech2gesture
- Owner: amirbar
- Created: 2019-04-03T02:14:33.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2024-03-06T19:07:58.000Z (10 months ago)
- Last Synced: 2024-12-30T00:12:09.098Z (14 days ago)
- Language: Python
- Homepage:
- Size: 1 MB
- Stars: 379
- Watchers: 27
- Forks: 44
- Open Issues: 12
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Learning Individual Styles of Conversational Gestures
#### [Shiry Ginosar](http://people.eecs.berkeley.edu/~shiry)* , [Amir Bar](http://amirbar.net)* , Gefen Kohavi, [Caroline Chan](https://www.csail.mit.edu/person/caroline-chan), [Andrew Owens](http://andrewowens.com/), [Jitendra Malik](https://people.eecs.berkeley.edu/~malik/)
##### Back to [main project page](https://people.eecs.berkeley.edu/~shiry/projects/speech2gesture/index.html)
## Prerequisites:
1. `python 2.7`
2. `cuda 9.0`
3. `cuDNN v7.6.2`
4. `sudo apt-get install ffmpeg`
5. `pip install -r requirments.txt`
## Data
1. Download the dataset as described [here](data/dataset.md)
## Instructions
1. Extract training/validation data
2. Train a model
3. Perform inference using a trained model
### Extract training data
Start by extracting training data:
```
python -m data.train_test_data_extraction.extract_data_for_training --base_dataset_path --speaker -np --speaker `
```
```
once done you should see the following directories structure:
(notice train.csv and a train folder within the relevant speaker)
Gestures
├── frames.csv
├── train.csv
├── almaram
├── frames
├── videos
├── keypoints_all
├── keypoints_simple
├── videos
└── train
...
└── shelly
├── frames
├── videos
├── keypoints_all
├── keypoints_simple
├── videos
└── train
```
`train.csv` is a csv file in which every row represents a single training sample. Unlike in `frames.csv`, here, a sample is few seconds long.
#### Columns documentation:
```
audio_fn - path to audio filename associated with training sample
dataset - train/dev/test
start - start time in the video
end - end time in the video
pose_fn - path to .npz file containing training sample
speaker - name of a speaker in the dataset
video_fn - name of the video file
```
### Training a speaker specific model
Training run command example:
```
python -m audio_to_multiple_pose_gan.train --gans 1 --name test_run --arch_g audio_to_pose_gans --arch_d pose_D --speaker oliver --output_path /tmp
```
During training, example outputs are saved in the define `output_path`
### Inference
optionally get a pretrained model [here](https://drive.google.com/file/d/1HpG9y7UHQ9VWJZnRxOW4O5_dOOBcBshR/view?usp=drive_link).
Perform inference on a random sample from validation set:
```
python -m audio_to_multiple_pose_gan.predict_to_videos --train_csv --seq_len 64 --output_path --checkpoint --speaker -ag audio_to_pose_gans --gans 1
```
Perform inference on an audio sample:
```
python -m audio_to_multiple_pose_gan.predict_audio --audio_path --output_path --checkpoint --speaker -ag audio_to_pose_gans --gans 1
```
### Reference
If you found this code useful, please cite the following paper:
```
@InProceedings{ginosar2019gestures,
author={S. Ginosar and A. Bar and G. Kohavi and C. Chan and A. Owens and J. Malik},
title = {Learning Individual Styles of Conversational Gesture},
booktitle = {Computer Vision and Pattern Recognition (CVPR)}
publisher = {IEEE},
year={2019},
month=jun
}
```