What is ArcGIS Reality Server
ArcGIS Reality Server is an ArcGIS Enterprise server role that provides distributed computing and storage for reality mapping. It allows you to manage and process large collections of overlapping drone, digital aerial, and satellite images and generate high-fidelity 2D and 3D products in a scalable, self-hosted environment. You can extend this capacity further with cloud computing resources. As a result, reality mapping tasks that used to take days or weeks can be completed in hours, and jobs that were previously too large to process can now be handled.
Key capabilities
ArcGIS Reality Server provides a comprehensive toolset for reality mapping workflows, including image management, adjustment, and generation of 2D and 3D products. These tools are based on the same technology in ArcGIS Pro but are optimized for enterprise scalability through distributed processing and project-based workflows.
Reality mapping workflow management
ArcGIS Reality Server uses projects and missions to manage reality mapping across one or more data capture sessions.
A project is a top-level workspace for managing reality mapping activities. It organizes all imagery acquisition, processing, and outputs associated with a specific geographic location or mapping effort across one or more missions.
A mission is a container for a single data capture session, holding all related imagery and the resulting processed data products. A project can include multiple missions from different capture sources, such as satellite, drone, or aerial. Each mission independently stores its own data products, which can include outputs such as DSM, true ortho, point cloud, mesh, and Gaussian splats, depending on the capture type and processing performed.

Block Adjustment
ArcGIS Reality Server includes the following advanced tools for image adjustment, which is essential for accurate geospatial products:
Tie Point Computation—Detect and match common points across overlapping images.
Triangulation—Calculate precise camera positions and orientations.
Orthorectification—Correct distortions caused by terrain and sensor geometry.
These tools use sensor-specific algorithms to ensure accuracy for different platforms, from UAVs to high-resolution satellites.
Generating products
ArcGIS Reality Server supports the creation of various 2D and 3D products from drone, digital aerial, and satellite images. The possible products are described below:
DSM—A digital surface model (DSM) of the earth and features above the surface. This is generated from a collection of overlapping images.
DTM—A digital terrain model (DTM) of the earth with above-ground objects such as trees, buildings, vehicles, and so on removed. A DTM is also referred to as bare earth elevation.
Orthomosaic—A composite image in which distortions caused by camera tilt, terrain, and lens distortion have been removed. This correction ensures that the scale is uniform across the image. It provides accurate locations for features on the ground, while above-ground features such as buildings and trees will appear to lean away from the camera and may not be in their correct horizontal location.
True Ortho—An orthorectified image without perspective distortion so that above-ground features do not lean and obscure other features.
DSM Mesh—A 2.5D textured model in which the adjusted images are draped on a triangulated irregular network (TIN) version of the DSM extracted from overlapping images.
Point Cloud—3D points representing the ground surface and features above the ground derived from overlapping imagery.
Mesh—A 3D textured model in which the ground and above-ground feature facades are densely and accurately reconstructed.
Gaussian Splats— A photo realistic 3D model created using high density points with an elliptical Gaussian distribution—called splats, instead of triangulated meshes used in 3D mesh generation. The splats are used with machine learning techniques to reconstruct high-fidelity 3D scenes that captures fine details like thin structures.
Note:
To generate Gaussian Splats, ensure all machines on the reality server site meet the minimum software and hardware requirements:
- At least 64 GB RAM.
- An NVIDIA GPU with at least 8 GB of VRAM and a compute capability of 7.5 or higher.
- CUDA 13 or later.
- Deep Learning Libraries installed.
Distributed processing
ArcGIS Reality Server uses parallel processing across multiple machines and distributed storage to handle many large-scale reality mapping tasks. This architecture reduces processing time, allowing organizations to complete projects in hours instead of days. Check this page to learn how to manage reality server project distribution for optimal performance.
Collaboration and sharing
ArcGIS Reality Server is designed for enterprise collaboration, as follows:
Members across your organization can contribute data, share expertise, and publish results securely.
All source data and generated products are stored and shared through your ArcGIS Enterprise portal, ensuring controlled access and governance.
Deployment
ArcGIS Reality Server can be deployed as follows:
On-premises, for organizations with strict data governance requirements.
In the cloud using Microsoft Azure. Deployment is streamlined with ArcGIS Enterprise Cloud Builder tools, which simplify site creation and configuration.
To deploy, use the standard ArcGIS Enterprise pattern to install the following:
The Enterprise portal, hosting ArcGIS Server site, and relational ArcGIS Data Store
ArcGIS Server on additional machines and assign the Reality Server role to the site
Learn more about deployment details and system requirements
Summary
ArcGIS Reality Server transforms imagery collections into high-fidelity 2D and 3D products quickly and securely by combining the following:
Distributed processing for speed
Enterprise collaboration for efficiency
This allows organizations to scale reality mapping workflows and deliver accurate, actionable geospatial intelligence for decision-making, planning, and digital twin initiatives.