This repository contains the model and supporting files to build a Docker container for autonomous segmentation of the temporal bone, intended for use with ABLInfer.
This model is available on DockerHub as uwoabl/temporal-bone-segmentation
.
- Docker, preferably with GPU support (tested and working on Windows WSL2 + CUDA)
- CPU with AVX (Intel ix-2xxx [Sandy Bridge, 2011], AMD Bulldozer [late 2011] or better)
- At least 26GB available RAM/swap (works fine with 16GB RAM + additional swap)
- If using a GPU, at least 6GB VRAM (more is better)
Building the image is generally a simple affair: change into the directory and run docker build -t <image-name> .
. Docker tends to make a bit of a mess with intermediate images; this can speed up future builds, but can also break things if the intermediate image that ran apt-get update
is out of date, since the build will try to retrieve non-existant packages. In this case, running docker build --no-cache -t <image-name> .
will solve the issue.