Home

Awesome

MLSLT: Towards Multilingual Sign Language Translation

This repository contains the SP-10 dataset described in "MLSLT: Towards Multilingual Sign Language Translation". The copyright of the data is owned by the non-profit association European Sign Language Centre, please ensure you have obtained permission before making any academic or commercial use.

Download videos

  1. Download repo.
git clone https://github.com/MLSLT/SP-10
  1. Install dependencies

If you use conda you can create a virtual environment with the following command,

conda env create -f env.yaml
conda activate sp-10

or you can use pip to install the corresponding packages.

pip install -r requirements.txt
  1. Install ffmpeg to process videos, if you use Ubuntu system you can use the following command to install.
sudo apt update
sudo apt install ffmpeg
  1. Download videos.
python video_downloader.py [--save_path ${SAVE_PATH}]

The dataset is saved in the path ./videos by default, and you can also use --save_path to customize the save path.

  1. Convert videos. If you customize the save path, please make sure that the save_path of this step is the same as the previous step.
python video_converter.py [--save_path ${SAVE_PATH}]

Folder structure

The storage structure of video data in folders is shown below.

.
├── dev
│   └── 635
│       ├── bg.mp4
│       ├── de.mp4
│       ├── en.mp4
│       ...
├── test
│   └── 1296
│       ├── bg.mp4
│       ├── de.mp4
│       ├── en.mp4
│       ...
└── train
    ├── 410
    │   ├── bg.mp4
    │   ├── de.mp4
    │   ├── en.mp4
        ...