Docker base image
When you package your Darcy AI applications into Docker containers, you need to include each of the software libraries that your application relies upon. Fortunately, we provide a Docker container base image that you can use when packaging your applications which contains each of the foundation components you need.
To add items to your Darcy AI application, you should include them in your own Dockerfile that builds on top of the base image. In some cases, however, you may need to alter the components found within the base image. The details of the Darcy AI Docker container base image are provided here so you can understand how it is built and alter it as needed.
Making your own base image should only be necessary in uncommon situations, such as when you need to specify a particular version of OpenCV because of compatibility issues with another library that you cannot substitute. If you must build your own base image, take note of the differences between ARM and x86 CPU platforms and revise the Dockerfile accordingly.
Structure of the base images
The Darcy AI container base image is built as a hierarchy. The image on the bottom is the foundation for all Darcy AI base images. The content of that image is described in the next section below.
A universal Darcy AI base image is made using the bottom image. It is universal because it includes the software components needed to run your Darcy AI application using either CPU processing or Google Coral processing.
Additionally, two other base images are made from the bottom image. One is for CPU processing only and the other is for Google Coral processing only. These base images are useful if you want to build your Darcy AI application from a smaller base image. This will result in a smaller total application image.
If you are not sure which base image is the best for your application, use the universal image which
is published as darcyai/darcyai.
You can see the actual base image Dockerfiles at the link below.
Darcy AI Docker base image files
Foundation base image
The Darcy AI container image on the bottom layer provides the largest amount of components. It is designed to install the correct version of many software libraries depending on the target CPU architecture for the build. You can see the actual Dockerfile at the following link.
There are a large number of software components being installed. To make the list easier to understand, here is a short list of the types of components found in this Dockerfile.
- SSL certificate libraries
- Network tools used to perform build operations
- Image and video codecs and libraries
- Compiler software used to build other components
- OpenCV, Pillow, and imutils Python libraries
When the target build architecture is ARM/v7, a large number of additional components must be installed. These are required for the Python libraries to work properly.
Universal base image
The universal base image is built on top of the bottom foundation base image. You can see the actual Dockerfile using the link below.
Universal base image Dockerfile
This base image is able to be used as the starting container image for your Darcy AI application packaging. It installs the Google Coral TPU software components which also includes the TensorFlow runtime that can be used by CPU processors.
This Dockerfile will continue to grow as support for additional types of AI processors is added to Darcy AI.
Google Coral base image
The Google Coral base image is also built on top of the bottom foundation base image, but it only contains the software components needed to run your Darcy AI application on devices that have a Coral TPU. You can see the actual Dockerfile using the link below.
Google Coral base image Dockerfile
CPU base image
The CPU base image is built on top of the bottom foundation base image, as well, and only contains the software components needed to run your Darcy AI application on devices with CPU processing. You can see the actual Dockerfile using the link below.