Video Features

video_features allows you to extract features from raw videos in parallel with multiple GPUs. It supports several extractors that capture visual appearance, optical flow, and audio features. See more details in Documentation.

Quick Start

or run with conda locally:

# clone the repo and change the working directory
git clone https://github.com/v-iashin/video_features.git
cd video_features

# install environment
conda env create -f conda_env_torch_zoo.yml

# load the environment
conda activate torch_zoo

# extract r(2+1)d features for the sample videos
python main.py \
    feature_type=r21d \
    device_ids="[0]" \
    video_paths="[./sample/v_ZNVhz7ctTq0.mp4, ./sample/v_GGSY1Qvo990.mp4]"

# use `device_ids="[0, 2]"` to run on 0th and 2nd devices in parallel
# or add `cpu=true` to use CPU

If you are more comfortable with Docker, there is a Docker image with a pre-installed environment that supports all models. Check out the Docker support. documentation page.

Supported models

Action Recognition

• I3D-Net RGB + Flow (Kinetics 400)
• R(2+1)d RGB (IG-65M, Kinetics 400)

Sound Recognition

• VGGish (AudioSet)

Optical Flow

• RAFT (FlyingChairs, FlyingThings3D, Sintel, KITTI)
• PWC-Net (Sintel)

Image Recognition

• ResNet-18,34,50,101,152 (ImageNet)

Language-Image Pretraining

• CLIP

Used in

• SpecVQGAN branch specvqgan
• BMT branch bmt
• MDVC branch mdvc

Please, let me know if you found this repo useful for your projects or papers.

Acknowledgements

• @Kamino666: added CLIP model as well as Windows and CPU support
• @ohjho: added support of 37-layer R(2+1)d favors.