OpenFace as docker container image for aarm64 environments.
The development of OpenFace stopped 5 years; nevertheless, it remains a valuable piece of software utilized in numerous research projects. OpenFace relies on outdated dependencies that pose challenges for compilation on aarm64 architecture. Therefore, I opted to develop a container that operates seamlessly on computers with ARM processors, such as the new Apple Silicon devices or AWS Graviton instances, eliminating the need of a complex compilation process and downloading large model files.
It requires to install Docker.
Process multiple images from a directory:
docker run --rm -it -v /host/input:/oface/input -v /host/processed:/oface/processed ghcr.io/juanparati/openface-aarm64:master ./FaceLandmarkImg -fdir input
Process a video file
docker run --rm -it -v /host/input:/oface/input -v /host/processed:/oface/processed ghcr.io/juanparati/openface-aarm64:master ./FaceLandmarkVid -f input/myvideo.mp4
Replace /host/input with the path to the local directory that contains the input files.
Replace /host/processed with the path to the local directory were the processed files are going to be generated.
- FaceLandmarkImg
- FaceLandmarkVid
- FaceLandmarkVidMulti
- FeatureExtraction
See origina/Dockerfile.