Video

Installation

If you would like to use the video augmentations, please install AugLy using the following command:

pip install augly[av]

This ensures that not only the base dependencies, but also the heavier dependencies required for audio & video processing, are installed.

In order to run the video tests and/or use the augmentations, you will also need to install ffmpeg. If you’re using conda you can do this with:

conda install -c conda-forge ffmpeg

If you aren’t using conda, you can run:

sudo add-apt-repository ppa:jonathonf/ffmpeg-4
apt install ffmpeg

AugLy Video uses distutils.spawn.find_executable() to automatically find your ffmpeg & ffprobe paths. If you would prefer to use another specific version of ffmpeg, export the AUGLY_FFMPEG_PATH and AUGLY_FFPROBE_PATH environment variables such that we can access the intended ffmpeg version:

which ffmpeg
export AUGLY_FFMPEG_PATH='<ffmpeg_path>'
which ffprobe
export AUGLY_FFPROBE_PATH='<ffprobe_path>'

Augmentations

Try running some AugLy video augmentations in Colab! For a full list of available augmentations, see here.

Our video augmentations use ffmpeg & OpenCV as their backend. All functions accept a path to the video to be augmented as input and output a file containing the augmented video. If no output path is specified, the original video input path will be overwritten. The file path to which the augmented video was written will also be returned by all augmentations.

Function-based

You can call the functional augmentations like so:

import augly.video as vidaugs

video_path = "your_vid_path.mp4"
output_path = "your_output_path.mp4"

vidaugs.add_dog_filter(video_path, output_path)
vidaugs.rotate(output_path, degrees=30)   # output_path will be overwritten

Class-based

We have also defined class-based versions of all augmentations, as well as a Compose operator used to combine transforms.

import augly.video as vidaugs

COLOR_JITTER_PARAMS = {
    "brightness_factor": 0.15,
    "contrast_factor": 1.3,
    "saturation_factor": 2.0,
}

AUGMENTATIONS = [
    vidaugs.ColorJitter(**COLOR_JITTER_PARAMS),
    vidaugs.HorizontalFlip(),
    vidaugs.OneOf(
        [
            vidaugs.RandomEmojiOverlay(),
            vidaugs.RandomIGFilter(),
            vidaugs.Shift(x_factor=0.25, y_factor=0.25),
        ]
    ),
]

TRANSFORMS = vidaugs.Compose(AUGMENTATIONS)

video_path = "your_vid_path.mp4"
out_video_path = "your_output_path.mp4"

TRANSFORMS(video_path, out_video_path)  # transformed video now stored in `out_video_path`

Unit Tests

You can run our video unit tests if you have cloned augly (see here) by running the following:

python -m unittest discover -s augly/tests/video_tests/ -p "*"

Note: some of the video tests take a while to run (up to a few minutes). If you want to run the 4 test suites individually, you can run any of the following commands (listed in order of increasing runtime):

python -m unittest augly.tests.video_tests.transforms.composite_test
python -m unittest augly.tests.video_tests.transforms.cv2_test
python -m unittest augly.tests.video_tests.transforms.ffmpeg_test
python -m unittest augly.tests.video_tests.transforms.image_based_test