Pretrained Deep Learning Models

A map of a neighborhood with houses outlined in red that uses pretrained deep learning models to identify homes

What are pretrained deep learning models?

ArcGIS pretrained deep learning models eliminate the need for huge volumes of training data, massive compute resources, and extensive artificial intelligence (AI) knowledge. Accelerate your geospatial workflows with built-in expertise and resources designed specifically for image feature extraction, land-cover classification, image redaction, and detecting objects. Automate the way you extract meaningful insights from imagery, point clouds, and video. 

What can pretrained models do for you?

  • Instantly extract features from a stream of data

    Pretrained deep learning models can instantly recognize complex shapes, patterns, and textures at various scales within images, point clouds, or video. This means that you can off-load tedious tasks of digitizing and extracting geographical features, such as roads, windows, and building footprints, with ease. 

  • Eliminate the need to develop and train models

    With pretrained models, you no longer have to invest time and energy into labeling datasets and training your own model. Our pretrained models are trained on huge volumes of data from a variety of geographies to learn what to look for. 

  • Retrainable to your geography and imagery

    Although our pretrained models aren't trained in all geographies, you can adapt them to your particular terrain, geography, and imagery type. Start with our models and then fine-tune the pretrained model to your needs using your labeled data. 

  • Available to ArcGIS users

    Access a growing number of pretrained deep learning models from ArcGIS Living Atlas of the World. The imagery team works hard to continually train and deliver new, complementary models to support your business needs. 

Previous
Next

Types of models

Pretrained deep learning models perform tasks, such as feature extraction, classification, redaction, detection, and tracking, to derive meaningful insights from large amounts of imagery. Solve problems for infrastructure planning and a variety of other applications.

How it works

Aerial view of a parking lot filled with vehicles representing object detection being applied

Choose your tool

Open your ArcGIS client application and launch your tool of choice (e.g., detect objects, classify objects, classify pixels).

Select your input

Point to the imagery you want to extract information from.

Extract

Browse for your desired pretrained deep learning model and run your analysis.

Video

Extracting railway assets from lidar from L&T

View Esri’s pretrained models at work with a demonstration of railway asset data extraction using LIDAR data and GeoAI from Larsen & Toubro (L&T).

Find your ArcGIS product

Pretrained models are designed to support and enhance your workflows regardless of your ArcGIS deployment. To run these models, you will need an ArcGIS Online, ArcGIS Pro, or ArcGIS Enterprise license along with the relevant extensions.

  • A purple outlined icon of a cloud converging with two square layers

    ArcGIS Online

    ArcGIS Image for ArcGIS Online is required to run these models. This extension offers a secure, scalable, and performant cloud environment to perform your analysis.

    Get ArcGIS Image for ArcGIS Online
  • A blue outlined icon of Earth

    ArcGIS Pro

    To extract image features, the ArcGIS Image Analyst extension is required. To extract point cloud features, you will need the ArcGIS 3D Analyst extension. 

    Get ArcGIS Image Analyst
    Get ArcGIS 3D Analyst
  • A purple outlined icon of a 3D box

    ArcGIS Enterprise

    Use the ArcGIS Image Server extension to process your models in ArcGIS Enterprise. Integrate machine learning and deep learning frameworks with ArcGIS Notebooks to solve spatial problems and perform image detection and classification.

    Get ArcGIS Image Server
    Explore deep learning in ArcGIS Notebooks
Previous
Next

Resources

Previous
Next

Explore all of our pretrained deep learning models

Browse models

Support

Customer service

Technical support

Give us a call

1-800-447-9778
Outside the United States