Sports Gesture Classification2: Build Dev Container

in Engineer on Project

Introduction to Dev Container

Dev container as Development Container on Vscode here, provide a self-featured development environment which is able to run an application on a separate tools, libraries from the local environement.

Features on Dev Container vs Docker Container

  • Common1: A separate environment with customization
  • Common2: Any changes made to the container will be lost when the container is stopped. However, the changes can be saved by committing the container to a new image. This new image can then be used to start a new container with the changes
  • Difference: Docker provides copy and mount features. However, the former does not synchronize changes between Docker and the local environment, and vice versa. In contrast, Dev Containers use a mount format for all files, ensuring that changes made locally or in Docker are synchronized on both platforms.

Set up Dev Container

Write Dockerfile

To set up a Dev Container, begin by creating a Dockerfile with content similar to what to find in the Sports Gesture Classification - Part1 Docker Container. And place it in the project directory.

Write Container Configuration File

Next, create a configuration file devcontainer.json at {project}/.devcontainer. The format of this file should resemble the following:

// README at: https://github.com/devcontainers/templates/tree/main/src/docker-existing-dockerfile
{
	"name": "Sports Gesture Classification Container",
	"build": {
		// Sets the run context to one level up instead of the .devcontainer folder.
		"context": "..",
		// Update the 'dockerFile' property if you aren't using the standard 'Dockerfile' filename.
		"dockerfile": "../Dockerfile"
	},
	"remoteUser": "linlin",
	"workspaceFolder": "/home/linlin/sport_gesture_classification_on_devcontainer",
	"mounts": [
		"source=/home/linlin/dataset/sports_kaggle/,target=/home/linlin/dataset/sports_kaggle,type=bind,consistency=cached",
	],
	"features": {
		"git": "latest"
	}
}

  • The default workspaceFolder is /home/{remoteUser}/workspace. Its property is to match the location of your mounted project in the Dev Container. This ensures that when you connect to the Dev Container, the terminal’s path will automatically be set to your project directory.

  • In mount property, the dataset is located at /home/linlin/dataset/sports_kaggle/ on the host machine and is mounted at /home/linlin/dataset/sports_kaggle inside the dev container.

    The mounts property is used to specify the location of the dataset on the host machine and the location where it should be mounted inside the dev container.

  • The type property is set to bind, which means that the dataset is mounted as a bind mount.

    This allows the dataset to be shared between the host machine and the dev container, and any changes made to the dataset inside the dev container will be reflected on the host machine.

  • The consistency property is set to cached

    It means that the dataset is mounted with cached consistency. This provides better performance when reading from the dataset.

Install Dev Container on Vscode

  1. On Extensions on Vscode, search and install ms-vscode-remote.remote-containers
  2. Ctrl + Shift + p, type Dev Container: Open Folder in Container then select current project folder. This action creates a Dev Container image and container.
    • The container and images would be listed on docker-desktop if it is installed.

Connect to the Dev Container on Vscode

  1. Click on Remote Explorer on left sidebar in Vscode

  2. Select on Dev Container on the top of Menu Bar

  3. Choose the current project container and click on -> to connect it
    • This opens a new Vscode window that operates within the Dev Container

  4. For rebuilding the container from this image, click on Run interactive on the image on Dev Images panel

Common Mistakes

It’s important to note that using docker build to create a Docker image, and then attempting to connect it under Dev Container in Vscode, may result in connection failures. Docker and Dev Containers are separate entities and cannot directly access each other’s systems.

Running Code on Dev Container

  1. On the new Vscode window, ctrl + r would pop up a terminal on vscode for running scripts inside the container
  2. Any modification on dev container would be synced up with the project locally
    • Git action can either be run on terminal or local

For details on training and API setup, you can refer to the instructions in Sports Gesture Classification - Part1 Docker.

Thank you for your attention and happy coding!