Docker images for computer vision and deep learning.
Moonbox are a set of Debian based Docker images made for training and evaluating deep neural networks on video input.
Python Base from a micromamba environment with basic compute libraries and runit-docker.
Build arg | Default | Description |
---|---|---|
baseimage |
mambaorg/micromamba:1.0.0-bullseye |
Set baseimage from which python base is built. |
docker build \
-t moonvision/python-base:latest \
-f docker/python-base/Dockerfile docker/python-base
Build arg | Default | Description |
---|---|---|
baseimage |
moonvision/python-base |
Set the baseimage from which moonbox is built. |
ffmpeg_from_docker |
moonvision/custom-builds:ffmpeg-4.2.1 |
The docker image from which FFmpeg will be installed. |
pytorch_from_docker |
moonvision/custom-builds:pytorch-1.11.0_torchvision-0.12.0 |
The docker image from which PyTorch and Torchvision will be installed. |
with_genicam |
false |
Whether to install GeniCam harvesters. |
with_cuda |
false / undefined |
If true install CUDA specific packages. If defined, set ENV variables to mount nvidia binaries at runtime. |
Moonbox Mini is the most lightweight Moonbox version. It includes Python, a couple of libraries and FFmpeg.
docker build \
-t moonvision/moonbox:mini-latest \
-f docker/moonbox/Dockerfile docker/moonbox
docker build \
-t moonvision/moonbox:genicam-latest \
--build-arg with_genicam=true \
-f docker/moonbox/Dockerfile docker/moonbox
docker build \
-t moonvision/moonbox:cuda-latest \
--build-arg with_cuda=true \
--build-arg ffmpeg_from_docker=moonvision/custom-builds:ffmpeg-cuda-4.2.1 \
-f docker/moonbox/Dockerfile docker/moonbox
A custom build of FFmpeg.
Build arg | Default | Description |
---|---|---|
baseimage |
mambaorg/micromamba:1-bullseye |
Set the image in which FFmpeg will be built. |
ffmpeg_version |
The FFmpeg release version to build. | |
with_cuda |
false |
Whether to build FFmpeg with Nvidia hardware acceleration. |
docker build \
-t moonvision/custom-builds:ffmpeg-latest \
--build-arg ffmpeg_version=4.2.1 \
-f docker/builders/ffmpeg/Dockerfile docker/builders/ffmpeg
docker build \
-t moonvision/custom-builds:ffmpeg-cuda-latest \
--build-arg ffmpeg_version=4.2.1 \
--build-arg with_cuda=true \
-f docker/builders/ffmpeg/Dockerfile docker/builders/ffmpeg
Build arg | Default | Description |
---|---|---|
baseimage |
moonvision/python-base:latest |
Set the image in which PyTorch will be built. |
pytorch_tag |
v1.11.0 |
PyTorch version (tag or commit if built, pip version if downloaded). |
torchvision_tag |
v0.12.0 |
Torchvision version (tag or commit if built, pip version if downloaded). |
prebuilt |
false |
If true download pip versions instead of building from source. |
with_cuda |
false |
Whether to build/download PyTorch with Nvidia hardware acceleration. |
docker build \
-t moonvision/custom-builds:pytorch-latest \
-f docker/builders/pytorch/Dockerfile docker/builders/pytorch
docker build \
-t moonvision/custom-builds:pytorch-cuda-latest \
--build-arg with_cuda=true \
--build-arg prebuilt=true \
-f docker/builders/pytorch/Dockerfile docker/builders/pytorch