How to Get Started Learning and Teaching Deep Learning Analysis in ArcGIS

09-16-2021 03:03 PM
Esri Contributor
2 0 370

The Esri Education team recently conducted the Demystifying Deep Learning Analysis for Your Students webinar. During the webinar, we received questions from the audience on:

  • How to get started?
  • Can I use the webinar contents to engage our students?
  • Can we get step-by-step instructions and where the data can be found?
  • Is Esri platforms offers deep learning tutorials (e.g Learn ArcGIS)?

The general answer to the questions above is YES, and below are further details pertaining to these questions.  

Webinar contents

The link to webinar slide deck and recording are available in this blog

Deep Learning workflow

At a high level, deep learning workflow in ArcGIS consists of the following steps:



In an essence, you need to label your data, train the deep learning model, and then run inferencing. But at a high level, it is the entire ecosystem working together. Begin with collections of imagery that may be managed by an image server (or cloud, or locally), and later use a suite of tools to perform downstream analysis to obtain the results. You can take it further by using apps like ArcGIS dashboard or ArcGIS StoryMap, to translate those results into actionable insights. This results in a complete end-to-end geospatial deep learning system.

Take a look of some deep learning applications to inspire you and your students. 


How to get started

As mentioned in the webinar, there are two options to get started with deep learning analysis.


Option 1: Use pre-trained AI models

This is the best way to introduce deep learning into your classroom. This option eliminates the need for having imagery for model training and labeling requirements, thus training the model itself (that may need massive computer requirements). It is both simple and time-saving.    

At the time this blog was posted, there were 20 ready-to-use pre-train models provided by Esri on the ArcGIS Living Atlas of the World. Take a tour of pre-trained AI models currently available in ArcGIS Living Atlas of the World to get a better idea of each model.  These pre-trained models can be used in ArcGIS Pro, ArcGIS Image Server and ArcGIS Image for ArcGIS Online. Each pre-trained model comes with a step-by-step PDF guide on how to use it. Watch the following short video to access and download this pre-trained model.


This option is the quickest and simplest option to introduce deep learning analysis for your students.


Option 2: Train your own model

The pre-trained models have been tested and work well in certain areas, especially globally.  However, there will be some cases that the models won’t perform well in your area of interest since the features/objects or other parameters may be different. In that case, you may prefer to utilize option 2: Train your own model, so it is specific to your geography area, resolutions, imagery properties and other asset types. ArcGIS has complete end-to-end capabilities to train your own models. You can use the tools available in ArcGIS Pro and ArcGIS Image for ArcGIS Online, or use ArcGIS Python API coding in notebooks. 

Resources to train your own model:

  • An article on how the deep learning models in the arcgis.learn module can be tapped into, to perform various GIS and remote-sensing tasks 
  • Watch for the upcoming releases of two Learn ArcGIS lessons on the topics:
    1. Automate Fire Damage Assessment (estimate release date October 2021)
    2. Classify Landsat 8 Imagery to a raster denoting Mangrove cover using deep learning in ArcGIS (estimate release date Nov 2021)


Installation of deep learning libraries

You need to install deep learning libraries to work with ArcGIS Pro and ArcGIS Enterprise.

To simplify the installation of dependencies of arcgis.learn, we provide in GitHub, deep learning libraries installers for ArcGIS (ArcGIS Pro, Enterprise, and ArcGIS API for Python). Install the package to get all the dependencies in place for your deep learning analysis.


More resources

GeoAI examples and resources in general:


Deep learning analysis is an exciting and valuable technology for students to learn. Wehope these resources can help you to get started. Feel free to contact me at if you have any questions, need some assistance, or would like to send us feedbacks for additional resources needed. 

About the Author
Canserina Kurnia is a GIS professional with over 18 years of experience. Currently holds the position of Solution Engineer with Esri Education team. Prior to that was a Solution Engineer and Technical Manager for Esri Global Asia Pacific, and Instructor/Team Lead with Esri Training Services