We recommend that you provision a Geo AI Data Science VM using the one-click provisioning solution template. However, if you prefer, you can create the VM from the Azure Portal using the instructions below:
- Log into the Azure Portal with your preferred Azure account.
- Click the "Create a resource" button at top left.
- In the marketplace search bar, type "Geo AI Data Science VM with ArcGIS" and press Enter. (Auto-completion will not work until the VM is publicly available.)
- Click on the corresponding result, then click "Create" to begin customizing your VM.
- In the Basics pane, provide your desired credentials, subscription, resource group, and deployment region, then click "OK."
- We recommend creating a new resource group, so that all associated resources can be easily deleted when you are done with the tutorial.
- If you have no preference, we recommend using the "East US" location: this will speed up transfer of tutorial files to your machine.
- On the Settings pane, choose an "NC6," "NC12," or "NC24" VM size to select a VM with 1, 2, or 4 GPUs, then click "OK".
- The tutorial can be completed with a single GPU. If you would like to explore the time savings achieved by training with multiple GPUs, you will need to select either an NC12 or NC24 VM size.
- If you do not see the "NC" VM sizes as options, you may need to choose a different location on the Basics pane.
- If you do not choose an "NC" VM size, you will likely encounter failures during model training later in this tutorial.
- Confirm your choices and buy the VM to begin deployment.
Deployment will likely take 10-20 minutes to complete. When it has finished, a link to your VM should be added to your dashboard in Azure Portal. (You can also find your VM by name using the search utility in Azure Portal.)
- Navigate to your VM's pane in Azure Portal.
- Along the top of the pane, you should see a "Connect" button. Click this to download an RDP file for connection to your VM via remote desktop.
- Your connection instructions will vary depending on your local machine's operating system:
- On Windows, double-click on the RDP file and supply the credentials you chose during provisioning to connect. (You may need to choose "Use a different account" and place a backslash in front of your username to override your default domain settings.)
- On Mac, you will need to install an application to connect using the RDP file. Further instructions are available here.
- Once you have connected to your VM, click on the "Jupyter" icon on the desktop to launch a Jupyter Notebook server.
- This will launch a new browser window or tab showing the Notebook Dashboard, a control panel that allows you to select which notebook to open.
- You may need to copy a URL from the auto-launched window into your preferred web browser in order to view the provided sample notebooks.
- Navigate to the "notebooks\GeoBooks\AI-for-Earth" subfolder and click on the
01_Intro_to_pixel-level_land_classification.ipynb
file to launch it.
This notebook will contain an explanation of the other notebooks in the subfolder, including a recommended order for reading and executing them. Note that code cells within the notebook can be executed by selecting the cell and pressing Ctrl+Enter.