/vrn-docker

:whale: Docker files for Large Pose 3D Face Reconstruction from a Single Image via Direct Volumetric CNN Regression

Primary LanguageLua

Docker build for vrn code

If you’ve been redirected here from the 3D face reconstruction demo website: Sorry, I turned it off.

This is intended to work on CPU. Will modify eventually to have CUDA support. It’s available on DockerHub but you are welcome to build it yourself too. If you feel like doing this yourself, please feel free to send a pull request to a branch called cuda.

Setup

Dockerhub

The pre-built docker image is available on Docker Hub and includes the CPU float versions of the models.

sudo docker pull asjackson/vrn:latest

Build docker manually

The model files were originally held in GitHub’s LFS, but I ran out of space. If you want to build the container yourself, feel free to ask me for links to these files, or alternatively grab them from inside the Docker Hub container.

git clone
cd vrn-docker
docker build -t vrn .

Running the container

An example image of Alan Turing is available in the folder data, this needs to be bind mounted to /data in the container. With any luck, data/turing.jpg.obj will be created.

docker run -v $PWD/data:/data:Z asjackson/vrn /runner/run.sh /data/turing.jpg

The Z suffix changes the SELinux context of the directory so that it can be accessed on systems which have SELinux enabled. This might not be necessary on some distributions, such as Ubuntu, where SELinux is disabled by default. However, it is likely required under CentOS and Fedora.