Development Environment Preparation

This page was machine-translated. The Japanese page is the official version.

Prepare Docker on your development PC and the hardware needed for testing ActDK applications. You will download and initialize ActDK in the next step, Setup ActDK.

Docker

Installation

Actdk internally uses Docker. See the document to install it.

Disable the containerd image store

Disable Docker’s containerd image store on the development PC where you use ActDK. If the containerd image store is enabled, actdk run --release may fail. This feature is enabled by default on fresh installations of Docker Engine 29 or later, so you need to disable it. If you upgraded from a version earlier than Docker Engine 29, the existing configuration remains in use unless you explicitly enable the containerd image store, so no change is needed.

No configuration changes are required on the remote device.

Docker Engine

Add the following configuration to /etc/docker/daemon.json on Linux. Create the file if it does not exist. If it already contains settings, preserve the other settings and set containerd-snapshotter to false under features.

{
	"features": {
		"containerd-snapshotter": false
	}
}

After saving the configuration, restart Docker. On systems using systemd, run the following command. Restarting Docker affects running containers, so finish your work before doing so.

$ sudo systemctl restart docker

Docker Desktop

In Docker Desktop, open Settings > General, disable Use containerd for pulling and storing images, and apply the settings.

After Changing the Configuration

Switching image stores hides the images and containers in the previous store. The data is not deleted and becomes visible again if you switch back. To use base images from the previous store in the new store, pull them from the registry again or rebuild them locally.

For details, see the official Docker documentation on the Docker Engine containerd image store and the Docker Desktop containerd image store.

Run as a non-privileged user

As actdk uses Docker internally, users must have permission to operate Docker. After a default installation, you need to be a privileged user.

See the document to run Docker as a non-privileged user.

Docker and QEMU

Docker supporting multi-architecture feature is required. When you install Docker, it is not enabled. Enable multi-architecture using multiarch/qemu-user-static as following:

$ docker run --rm --privileged multiarch/qemu-user-static --reset -p yes

For more details, see Multi-arch All The Things for details.

Required Hardware

ActDK assumes the following two-stage development cycle:

Develop an application with a PC and a Raspberry Pi by using a direct connection.
Upload the application to Actcast, and test the online-built application with a Raspberry Pi.

First, develop an Actcast application on your PC and communicate it directly with the Raspberry Pi to check its operation. You can check the operation quickly because the online build explained later is not performed, but ActDK Overview as mentioned above, deep learning models are transformed into slower implementations.

After confirming the operation in the local environment, the next step is to check the operation in the same way as the finally released application. Upload your application to Actcast; it will re-create your application with faster implementation of your deep-learning model.

A Raspberry Pi and a micro SD card are required for local testing in development mode.

To test the application you create in this tutorial, you also need an official Raspberry Pi Camera Module.

See also Getting Started with Actcast for a list of required devices.

Previous: ActDK Overview

Next: Setup ActDK

Last updated on 2026/11/13