RealityScan 2.2 Adds AMD GPU Support and New 360 Camera Workflow

RealityScan 2.2 is now available with support for AMD GPUs, a new workflow for processing 360 camera images, and multiple fixes across alignment, texturing, editor tools, and command-line operations.

This update removes the NVIDIA-only requirement for GPU acceleration and expands hardware options for RealityScan users running photogrammetry and reality capture projects.

---

AMD GPU Support Now Available in RealityScan 2.2

RealityScan 2.2 adds support for AMD Radeon and Radeon PRO GPUs on Windows.

Previously, GPU-accelerated reconstruction was only available on NVIDIA hardware. With this release, AMD GPUs can now accelerate the same reconstruction stages supported on NVIDIA GeForce and Quadro GPUs.

RealityScan also supports mixed-GPU systems. Users can install both AMD and NVIDIA GPUs in the same workstation, and RealityScan will distribute reconstruction workloads across all supported GPUs simultaneously.

---

Supported AMD GPUs

RDNA 3 (gfx1100)

• Radeon RX 7900 XTX
• Radeon RX 7900 XT
• Radeon PRO W7900
• Radeon PRO W7800

RDNA 3 (gfx1101)

• Radeon RX 7800 XT
• Radeon RX 7700 XT
• Radeon PRO W7700

RDNA 3 (gfx1102)

• Radeon RX 7600 XT
• Radeon RX 7600
• Radeon RX 7650 GRE

RDNA 3.5 (gfx1151)

• Ryzen AI Max Series APUs (Strix Halo)

RDNA 4 (gfx1200)

• Radeon RX 9060 XT
• Radeon RX 9060

RDNA 4 (gfx1201)

• Radeon RX 9070 XT
• Radeon RX 9070
• Radeon RX 9070 GRE
• Radeon AI PRO R9700

AMD GPU support is currently available on Windows. Linux support is planned for a future release.

---

Using 360 Cameras in RealityScan

RealityScan 2.2 introduces a documented workflow for processing images captured with 360-degree cameras.

A 360 camera can capture an entire environment in a single image, making it useful for scanning interiors, large spaces, and environments where capturing traditional photos would take more time.

Most 360 cameras output images in equirectangular format. These images contain a full 360-degree horizontal view and a 180-degree vertical view of the scene.

Because photogrammetry workflows typically use standard perspective images, RealityScan recommends converting equirectangular images into cube-face views before processing.

---

Converting Equirectangular Images to Cube Maps

RealityScan provides a web-based tool that converts equirectangular images into cube-face images that can be imported into a project.

The conversion process generates:

• Cube-face images for reconstruction
• Camera calibration information
• Focal length data
• XMP metadata files

The generated metadata helps RealityScan align images more accurately because the software already knows the camera properties and field of view.

Since equirectangular images are already undistorted, RealityScan can skip certain correction steps when accurate metadata is available.

---

Cube Map Settings

Field of View (FOV)

The default cube-map configuration uses a 90-degree field of view for each face.

Users can increase the FOV up to 120 degrees.

A wider field of view creates overlap between cube faces, which can improve image alignment because more visual information is shared between adjacent images.

Face Size

The conversion tool automatically calculates the recommended face size based on the selected field of view.

Users can manually specify a face size if they need to reduce storage requirements.

XMP Metadata Files

The conversion tool can generate XMP sidecar files containing camera metadata.

These files provide information such as:

• Camera calibration
• Focal length
• Field of view

When imported with the images, this data can improve alignment accuracy and reduce processing time.

---

Recommended Capture Settings for 360 Cameras

RealityScan provides several recommendations for capturing images with 360 cameras.

Use the Highest Available Resolution

Higher-resolution images contain more visual detail and provide more information for reconstruction.

Capture Multiple Passes

Make multiple passes through the area being scanned rather than relying on a single walkthrough.

Additional passes increase coverage and improve alignment opportunities.

Change Camera Height

Capture the scene from noticeably different heights.

The variation creates parallax, which helps RealityScan determine spatial relationships between images.

Maintain Strong Image Overlap

A starting target of 70% to 80% overlap between frames is recommended.

If alignment issues occur, increase frame density by capturing more images.

If alignment is successful but processing takes longer than necessary, image density can be reduced.

Adjust Frame Density Based on the Environment

Smaller rooms and environments with nearby objects require more frequent image captures because the scene changes more quickly as the camera moves.

Large open spaces generally require fewer images because the scene changes more slowly.

Walking speed also affects image spacing. Moving too quickly through a dense environment can reduce overlap and make alignment more difficult.

---

Editor Fixes

RealityScan 2.2 includes several fixes for editor-related issues.

Fixed issues include:

• An "end of physical tape" error when importing COLMAP data from the 3D Maker Pro Eagle SLAM scanner
• An "end of physical tape" error caused by trailing spaces in imported COLMAP data
• A crash that could occur when using Marginal Triangle Selection and Set Reconstruction Region tools in a specific sequence
• A problem where camera selections in the Render Image dialog were not applied during export

---

Alignment Fixes

Several alignment-related issues have been resolved.

Fixes include:

• Saved Trajectory settings generating invalid configuration files
• Incorrect camera rotation in sparse reconstruction reports exported for web viewing
• Alignment processes hanging indefinitely on certain datasets
• A crash that could occur when using locked camera positions
• Alignment generating a component containing only a single camera

---

Texturing Fixes

RealityScan 2.2 fixes an issue where reprojecting textures between two planes could create visible pattern artifacts in displacement maps.

---

Command-Line Interface (CLI) Fixes

The update also includes fixes for command-line workflows.

Resolved issues include:

• The -addImageWithCalibration command not applying proposed XMP metadata when importing an image already present in the project
• The -importLaserScanFolder command incorrectly placing .lsp files during import

---

RealityScan 2.2 Summary

RealityScan 2.2 introduces AMD GPU support, allowing Radeon and Radeon PRO users to access GPU-accelerated reconstruction workflows previously limited to NVIDIA hardware. The release also adds a documented workflow for using 360 camera imagery in photogrammetry projects and includes fixes across alignment, texturing, editor tools, and CLI operations.

Users running RealityScan on Windows can download the update now, while Linux support for AMD GPUs is planned for a future release.

---

Get RealityScan Through Motion Media

Motion Media is an authorized RealityScan reseller serving creative professionals, visualization teams, surveyors, and reality capture specialists.

If you're looking for RealityScan licenses, upgrades, or deployment guidance, contact Motion Media for pricing and product assistance.