Get ready for deep learning in ArcGIS Pro

Deep learning is used by ArcGIS Pro tools to solve spatial problems, detect objects, and perform pixel classification. Using these tools requires that you have the correct deep learning libraries installed on your computer. In this tutorial, you will learn how to get ready for deep learning, setting up the libraries and checking that the installation was successful. Optionally, you will also learn how to verify your computer’s settings and troubleshoot common issues that might occur.

This tutorial was last tested on November 22, 2023, with ArcGIS Pro 3.2. If you're using a different version of ArcGIS Pro, you may encounter different functionality and results.

Requirements

• ArcGIS Pro (see options for software access)
• ArcGIS Image Analyst
• Recommended: NVIDIA GPU with a minimum of 8 GB of dedicated memory.
• About 10 GB of disk space, plus an extra 6 GB to download and extract temporary files.

Set up ArcGIS Pro for deep learning

ArcGIS Pro deep learning tools use Python deep learning libraries behind the scenes. You’ll install the libraries and verify their successful installation.

Install deep learning libraries

First, you’ll download the deep learning libraries and install them.

Caution:

If, in the past, you already installed the deep learning libraries for an older version of ArcGIS Pro, see the Update if you already installed the libraries for an older version of ArcGIS Pro section later in the tutorial for more specific instructions.

1. Ensure that the ArcGIS Pro application is closed.
2. Go to the Deep Learning Libraries Installers for ArcGIS page.
3. Under Download, click Deep Learning Libraries Installer for ArcGIS Pro 3.2.

   

   Note:

   If you are using a previous release of ArcGIS Pro, you can find the corresponding libraries by clicking Downloads for previous versions.

   If you are using a recently released version of ArcGIS Pro, the deep learning libraries for that new version will be available on the same day or within a couple of days of the release.

   The ArcGIS_Pro_32_Deep_Learning_Libraries.zip compressed folder downloads to your computer.

4. In Microsoft File Explorer, go to the Downloads folder and locate the ArcGIS_Pro_32_Deep_Learning_Libraries.zip file.
5. Right-click ArcGIS_Pro_32_Deep_Learning_Libraries.zip and extract it to a folder on your computer.

   You can do that by choosing the Extract All menu option or another archive extraction application, such as 7-Zip.

   

6. Open the ArcGIS_Pro_32_Deep_Learning_Libraries extracted folder.

   

   Caution:

   If you attempt to run the ProDeepLearning.msi installer package from within the compressed .zip file instead of the extracted folder, the installation process will fail.

7. Double-click the ProDeepLearning.msi Windows installer package to install the deep learning libraries.

   

   Note:

   If you don’t see the file extensions, you can turn them on from the View tab and in the Show/hide section.

   The installer window appears.

   

8. In the installer window, click Next twice.
9. Click I accept the master agreement, and click Next.
10. Click Install.
11. When the message Do you want to allow this app to make changes to your device? appears, click Yes.

    The deep learning libraries install on your computer.

    Note:

    The installation may take several minutes to complete.

12. When the installation is complete, click Finish.

Verify the installation is successful

Now that you have installed the deep learning libraries, you'll open ArcGIS Pro and verify that the installation was successful.

1. Start ArcGIS Pro. If prompted, sign in using your licensed ArcGIS organizational account.

   Note:

   If you don't have access to ArcGIS Pro or an ArcGIS organizational account, see options for software access.

2. Click Settings.

   

3. Click Package Manager.

   

   In the Installed Packages section, you can see the packages that you installed. These are used behind the scenes to run analyses.

   Note:

   If you previously created another Python environment for ArcGIS Pro, you will first need to switch to the default arcgispro-py3 environment. The deep learning libraries are automatically installed in arcgispro-py3.

4. In the Installed Packages section, check that the deep-learning-essentials package is listed.

   It is one of the required packages for performing deep learning training and inferencing in ArcGIS Pro.

   

You installed the deep learning libraries and confirmed their installation in ArcGIS Pro. You are now ready to perform the deep learning analyses presented in the Learn ArcGIS tutorials on deep learning or other deep learning workflows of your choice in ArcGIS Pro.

As you run the deep learning tools, you may occasionally encounter some issues, based on your computer system configurations. If that’s the case, pause your work and explore the next part of this tutorial, Troubleshoot common issues, to learn more.

Note that the deep learning tools in ArcGIS Pro may mention the choice between using CPU or GPU processors. If you are not sure about the options available on your computer, read the Check for GPU availability section.

Troubleshoot common issues (optional)

Installing the deep learning libraries as you did previously is the most important step to get ready for deep learning in ArcGIS Pro, and it is usually all you need. However, in some cases, you may need to take some extra steps to ensure that your computer system is configured properly and will be successful at running deep learning workflows. If you want to learn more or if you are experiencing issues while running deep learning tools in ArcGIS Pro, check out the following tips.

Check for GPU availability

When running deep learning tools in ArcGIS Pro, it is recommended that you run them using your computer’s GPU (Graphics Processing Unit) instead of your computer’s standard processor, the CPU (Central Processing Unit). Deep learning techniques require a lot of computing power, and GPUs are well adapted because they can perform multiple, simultaneous computations. However, not all GPUs will work for deep learning. More specifically, the deep learning libraries expect an NVIDIA GPU platform. In this section, you will learn how to check whether your computer has a suitable GPU.

Note:

If your computer has only a CPU, you can still use it for simpler deep learning workflows (as illustrated, for instance, in the Detect objects with a deep learning pretrained model tutorial), but the processing time will be longer.

To know whether your computer has NVIDIA GPU capabilities, perform the following steps.

1. In the Windows taskbar, click the Search button.

   

2. In the search box, type Task Manager and press Enter.

   

3. In the Task Manager window, click the Performance tab.

   

4. In the list of performance indicators, look for a GPU option of the NVIDIA type.

   

   • If an NVIDIA GPU is listed, you can run ArcGIS Pro deep learning tools in GPU mode, and you’ll obtain a faster performance.
   • If there is no mention of NVIDIA GPU, your computer doesn’t have that capability. If asked to choose, you should run ArcGIS deep learning tools in CPU mode.
   • Your computer might have two GPUs, one NVIDIA and the other one non-NVIDIA. When running a deep learning tool in GPU mode, ArcGIS Pro will automatically select the NVIDIA GPU and will ignore the other one.

   

   An example shows a computer with two GPUs, including an NVIDIA one.

   Note:

   To learn more, read how to check your GPU and why it matters.

   If you have an older computer, you may have an NVIDIA GPU that won’t work with the deep learning libraries. Learn more about your NVIDIA GPU compute capabilities.

   The main indicator to measure your NVIDIA GPU’s power is the amount of dedicated memory. 4 to 8 GB of dedicated memory is recommended: 4 GB if running inferencing only, and 8 GB for training deep learning models from scratch. You will now check the specifications of your NVIDIA GPU.

5. Click the NVIDIA GPU indicator to display more information about it.
6. Locate the Dedicated GPU memory indicator.
   • The second number indicates the total amount of dedicated memory available (8 GB in the following example image).
   • The first number indicates how much dedicated memory is currently in use (1.9 GB in the example image).

   

7. Locate the Dedicated GPU memory usage graph.

   

   This graph shows the dedicated GPU memory usage in real time. When running a deep learning tool in ArcGIS Pro, you will see the usage level shoot up on this graph.

8. Examine the rest of the information provided.

   The following specifications are of particular interest:

   • The name of the GPU model (in the following example image, NVIDIA Quadro RTX 4000).
   • The Driver version (here 31.0.15.3598).
   • The Driver date (here 5/24/2023).

   

This information will be crucial if you encounter issues due to your GPU driver not being up to date. You will learn more about this situation in the next section, Update your NVIDIA driver.

Update your NVIDIA driver

An out-of-date GPU driver will cause deep learning tools to fail. If you encounter this issue, here is how to update your GPU driver.

1. Ensure that the ArcGIS Pro application is closed.
2. Ensure that you collected the information about your NVIDIA GPU’s model and driver version (see the Check for GPU availability section earlier in this tutorial).
3. Go to NVIDIA’s Download Drivers page.
4. In the NVIDIA Driver Downloads section, choose the options that match your GPU. In this example, the GPU model is NVIDIA Quadro RTX 4000.

   • 

5. Click Search, click Download, and click Agree & Download.

   The driver’s installer file downloads to your computer.

6. On your computer, locate the installer file (usually in your Download folder) and run it.
7. Click Yes to allow changes to the system and click OK to run the installer.
8. In the NVIDIA Installer window, click Agree and Continue, accept the default installation options, and click Next.
9. When the installation is finished, click Close.

Your GPU driver is now up to date.

Update if you already installed the libraries for an older version of ArcGIS Pro

If you installed the deep learning libraries on an older version of ArcGIS Pro, and then subsequently upgraded ArcGIS Pro to a more recent release, the version of the libraries will not match your geoprocessing tools and may generate errors. You can check the version of the deep learning libraries currently installed on your computer using the following steps.

1. In the Windows taskbar, click Search.

   

2. In the search box, type Add or remove programs and press Enter.

   

   The Add or remove programs on your computer window appears.

3. In the search box, type deep.

   

4. In the list of results, click Deep Learning Libraries for ArcGIS Pro.

   

   The version number of the libraries appears.

   

   The first two digits (in this case, 3.2) indicate the corresponding ArcGIS Pro version. If you have ArcGIS Pro 3.2 installed, ArcGIS Pro and the deep learning libraries are in sync. If, instead, you find that the deep learning libraries version is older than your ArcGIS Pro version, there is a mismatch, and the deep learning tools will not work correctly.

5. Close the Add or remove programs on your computer window.

If you found a version mismatch, you must remedy the issue by following the instructions on the Deep Learning Libraries Installers for ArcGIS page under the Upgrading From a Previous Version section.

Manage environments with deep learning libraries

In ArcGIS Pro, Python is run by default within a Python environment named arcgispro-py3. In some cases, you may have created a separate environment to install different Python libraries. The deep learning libraries installer will always install the libraries in the default arcgispro-py3 environment. If you had switched to an environment other than arcgispro-py3, you will need to switch back to arcgispro-py3 before you can use the deep learning tools. You can do that with the following steps.

1. Start ArcGIS Pro.
2. Click Settings.

   

3. Click Package Manager.

   • 

4. For Active Environment, review the environment name selected.

   In the following example image, the current active environment is myenv, a custom Python environment created by the user.

   

   You will activate the default environment instead.

5. Click the drop-down arrow and choose arcgispro-py3.

The default Python environment, arcgispro-py3, is now activated. This is where the deep learning libraries were installed. You will now be able to use ArcGIS Pro deep learning tools successfully.

Clone the default environment with or without the deep learning libraries

To create a cloned ArcGIS Pro environment that contains the deep learning libraries, you can clone the arcgispro-py3 default environment after you’ve installed the libraries. The deep learning libraries will be included in the new environment.

In some other cases, you may need an environment that does not include the deep learning libraries, for example, if you attempt to install another Python library and find that some packages conflict. You can create such an environment by using the Python command prompt, as explained in the following instructions.

1. Exit ArcGIS Pro.
2. In the Windows taskbar, click the Search button.

   

3. In the search box, type Python Command Prompt and press Enter.

   

4. In the Python Command Prompt window, type the following Conda command:

   conda create -n myenv arcpy=3.2

   In this example, myenv is the name of the new environment that you are creating, and 3.2 is the version of the ArcPy package, which matches the ArcGIS Pro version installed on your computer (in this case ArcGIS Pro 3.2).

   

   Note:

   You can replace myenv with a name of your choice.

5. Press Enter.

   The command starts running.

6. If asked Proceed ([y]/n), type y and press Enter.

   

   Note:

   The process may take several minutes to complete.

7. When the process is complete, start ArcGIS Pro.
8. Click Settings.
9. Click Package Manager.
10. For Active Environment, review the environment name selected.

    In the following example image, the default environment, arcgispro-py3, is currently active.

    

    You will activate the environment you just created.

11. Click the drop-down arrow and choose myenv (or the other name you chose for your custom environment).

    

12. Use the Add Packages tab to install the other (non-deep learning) packages that you want to use in the new environment.

    

    Note:

    Learn more about Package Manager and how to add packages.

Your new custom Python environment is now active. It doesn’t contain the deep learning libraries, but it contains the other (non-deep learning) Python packages you need.

Memory problems

When running a deep learning tool in ArcGIS Pro, it may run out of dedicated GPU memory and fail. This is generally because the Batch Size value you chose is too large for your GPU. To resolve this, you can reduce the Batch size value (for instance, from 4 to 2 or even 1) and run the tool again.

Reporting issues and asking questions

There are several ways of reporting issues or asking questions concerning deep learning in ArcGIS Pro.

• If you encounter unexpected behavior while installing the deep learning libraries, you can report the issue to the Deep Learning Frameworks GitHub repository .
• You can post your questions about using deep learning tools in ArcGIS Pro in the ArcGIS Image Analyst section of Esri Community. You can also post these questions to the ArcGIS Pro Questions section, adding the tag deep learning.
• The deep learning tools are fully supported, so you can also open a support case.

Additional resources

To learn more about the topics presented in this tutorial, check the following additional resources.

• Install deep learning frameworks for ArcGIS
• Deep learning frequently asked questions
• ArcGIS Pro 3.2 system requirements
• Deep Learning with ArcGIS Pro Tips & Tricks: Part 1

In this tutorial, you learned how to get ready for deep learning in ArcGIS Pro. You set up the Python deep learning libraries and checked that the installation was successful. You then learned how to verify your computer’s settings and troubleshoot common issues that might occur. You are now all set to run deep learning workflows in ArcGIS Pro.

You can find more tutorials in the tutorial gallery.

Acknowledgements

• Gallery and banner images: © 2023 Adobe Stock. All rights reserved.