signlang-augmenter

signlang-augmenter is a Python library for sign language video augmentation.

The first shipped domain is video augmentation. The current video API exposes 44 supported augmentation keywords across 5 groups: spatial, photometric, noise/quality, temporal, and camera simulation. Pose augmentation is planned for a later phase and the pose package remains a placeholder for that future work.

What is included

• Video I/O helpers for reading and writing clips.
• Reproducible augmentation helpers with a shared seed function.
• A single convenience entry point for keyword-based usage: augment_video.
• Separate augmentation modules for spatial, photometric, noise/quality, temporal, and camera simulation transforms.
• A public API under signlang_augmenter.video with 11 exported helpers.

Install

Option 1: Install the library to use it

pip install signlang-augmenter

Or install it directly from GitHub:

python -m pip install "git+https://github.com/24-mohamedyehia/signlang-augmenter.git"

Option 2: Install the project for development

If you want to modify the code and contribute, use an isolated environment and editable install:

1. Install Miniconda if needed.
2. Run:

git clone https://github.com/24-mohamedyehia/signlang-augmenter.git
cd signlang-augmenter
conda create -n signlang-augmenter python=3.11 -y
conda activate signlang-augmenter
python -m pip install --upgrade pip
python -m pip install -e ".[dev]"

Quick Import Check

Use this when you want a fast sanity check that the library imports correctly and the public entry point is available:

from signlang_augmenter.video import augment_video, list_supported_video_methods

print(len(list_supported_video_methods()))
print(list_supported_video_methods())

If the import succeeds, the library is ready to use.

Example

Open examples/basic_pipeline.ipynb for a notebook that shows the video-only workflow with one convenience-wrapper example and separate examples for each augmentation group.

Typical usage:

from pathlib import Path

from signlang_augmenter.video import augment_video

input_video = Path("./data/sample.mp4")
output_dir = Path("./outputs")

outputs = augment_video(
    input_video_path=input_video,
    output_dir=output_dir,
    speed_up=1.15,
    frame_sampling=0.75,
    random_crop=0.9,
    horizontal_flip=0.3,
)

Public API

Use signlang_augmenter.video as the main entry point. It exports 11 helpers:

• augment_video
• list_supported_video_methods
• set_seed
• read_video_cv2
• write_video_cv2
• spatial_augmentations_to_videos
• photometric_augmentations_to_videos
• list_supported_photometric_methods
• noise_quality_augmentations_to_videos
• temporal_augmentations_to_videos
• camera_simulation_to_videos

For the full breakdown of categories, defaults, and method counts, see docs/README.md.