Dockerfile to create a Docker container with the COLMAP executables.
The container is hosted on Docker Hub. It is built from a CUDA image, and therefore requires both the NVIDIA Docker plugin and a CUDA supported graphics card. To use it, simply execute
# Only CPU
docker run --rm -v $PROJECT_PATH:/project gerardobort/colmap colmap automatic_reconstructor --workspace_path /project/tmp --image_path /project/bunny --use_gpu=0
# Leveraging CUDA GPUs
nvidia-docker run --rm -v $PROJECT_PATH:/project gerardobort/colmap colmap automatic_reconstructor --workspace_path /project --image_path /project/images
You must mount a volume to the container using the -v flag in order to provide input images to COLMAP as well as to access the output. In the example above, the $PROJECT_PATH environment variable should be set to point to a directory on your host computer containing your project files.
See the COLMAP documentation for instructions on how to use the COLMAP executables.
This Dockerfile is based on Greg Anders' colmap-docker project.