slt_datasets 0.0.2rc45.post1

Multilanguage datasets for Sign Language Translation.

SLT Datasets Downloader

Overview

SLT Datasets Downloader is a Python library that allows users to download and process multiple sign language translation (SLT) datasets from different languages. It is designed to facilitate the training of machine learning models for SLT tasks.

Features

• Supports multiple sign language datasets.
• Provides tools for downloading, preprocessing, and tokenizing datasets.
• Compatible with PyTorch for model training.
• Handles different input types (video, pose) and output types (text, gloss).

Supported Datasets

The following datasets are supported:

Name	Input Language	Target Language	Status	# Samples	Hs	Video	Pose	Transcription	Gloss	Other data	# Words	# Singletons	# Signers	Source	BLEU (4)
RWTH-PHOENIX Weather 2014 T	GSL	German	Downloaded	7,096	10	Yes	No	Yes	Yes	-	2,887	1,077	9	TV	28.95
LSA-T	LSA	Spanish	Downloaded	14,880	22	Yes	Yes (MP)	Yes	No	-	14,239	7,150	103	Web	0.93 (WER)
How2Sign	ASL	English	Downloaded	35,000	80	Yes	Yes	Yes	Yes	Multiple angles, 3D pose, Speech	16,000	-	11	-	8.03
LSFB-CONT	French-Belgian SL	French-Belgian Glosses	Downloaded	85,000	25	Yes	Yes (MP)	No	Yes	-	6,883	-	100	Laboratory	-
ISLTranslate	Indian SL	English	Downloaded	31,222	-	Yes	Yes (MP)	Yes	No	-	11,655	-	-	-	6.09
GSL	Greek SL (GSL)	Greek	Downloaded	40,826	10	Yes	No	No	Yes	Depth	310	0	7	Laboratory	20.62 (WER)

Installation

To install the library and its dependencies, run:

pip install -r requirements.txt

Usage

Loading a Dataset

from slt_datasets.SLTDataset import SLTDataset

dataset = SLTDataset(
    data_dir="/path/to/dataset",
    input_mode="video",
    output_mode="text",
    split="train"
)

Accessing Samples

sample_input, sample_output = dataset[0]

Visualizing Pose Data

dataset.visualize_pose(0)

Project Structure

 |-src
 | |-slt_datasets
 | | |-SLTDataset.py  # Main dataset loader
 | | |-WordLevelTokenizer.py  # Tokenizer for text processing
 | | |-dataset_comparison.ipynb  # Notebook for dataset comparison
 |-tests
 | |-test_methods.py  # Unit tests for the dataset loader
 |-docs  # Documentation files
 |-requirements.txt  # Dependency list
 |-pyproject.toml  # Project configuration
 |-README.md  # This file

Contributing

Contributions are welcome! Please follow the guidelines in CONTRIBUTING.md and ensure your code passes all tests before submitting a pull request.

License

This project is licensed under the terms of the LICENSE file.

Support

For any issues or questions, please refer to SUPPORT.md or open an issue in the repository.