Automate 2026: RealSense Debuts AI-Powered Vision for Robotics

The company unveiled the RealSense D585 Pro depth camera featuring integrated AI processing and the Perception Studio beta program offering early access to new features within the RealSense ecosystem.


News 2 hours ago by Seth Price

RealSense unveiled two new AI-enhanced vision solutions at Automate 2026. The RealSense D585 Pro is a system-on-chip (SoC) depth camera designed for industrial robots, AMRs, and inspection systems. The company also released the Perception Studio beta program, which introduces new capabilities within the RealSense 2.0 SDK environment.

 

The D585 Pro features a 120° × 100° field of view at 60 fps, as well as an on-device AI inference engine

The D585 Pro features a 120° × 100° field of view at 60 fps, as well as an on-device AI inference engine. Image used courtesy of RealSense

 

D585 Pro AI-Native Vision

The RealSense D585 Pro is not simply a new and improved camera; it is a depth camera redesigned from the ground up to meet the needs of robotics and inspection tasks. Compared to previous RealSense models, it provides twice the depth quality.

In terms of viewing capability, the D585 Pro can perform high-resolution (1280 x 960) imaging at a standoff distance of under 15 cm, all at 60 fps. This enables the D585 Pro to capture more details at higher speeds, increasing throughput on a quality assurance line or providing higher-quality data for autonomous navigation.

Additionally, its 120° x 100° field of view is designed specifically for SLAM applications. AMRs with this camera will get the “full picture” of their environment, not just a narrow snapshot.

 

The D585 Pro brings AI-powered vision to robotic applications, including humanoids and AMRs

The D585 Pro brings AI-powered vision to robotic applications, including humanoids and AMRs. Image used courtesy of RealSense

 

On-Device AI Processing

It would be a mistake to think of the D585 Pro camera as a close-range-only type camera. Standoff distances can reach as far as 10 m. Not only that, it has several built-in IR filters that allow the camera to be used in direct sunlight or in low-light indoor applications. Furthermore, the camera is sealed to IP65 standards, allowing it to be used indoors or outdoors, as well as in challenging industrial environments.

At the heart of the D585 Pro is a proprietary, fifth-generation SoC design that contains a quad-core ARM processor. This processor has built-in image sensing capability, a digital signal processor, and AI-enhanced accelerators that make high-resolution imaging possible. Ultimately, this camera is not just a sensor but a fully capable imaging system for edge computing.

 

Perception Studio includes three key features: visual-inertial odometry, close-range depth, and person detection

Perception Studio includes three key features: visual-inertial odometry, close-range depth, and person detection. Image used courtesy of RealSense

 

Perception Studio Beta Program

RealSense also released a beta version of its Perception Studio. Perception Studio is a software package that extends the capabilities of the existing RealSense 2.0 SDK.

The beta features include improved sensing for the D400 line of sensors, human detection for SLAM and human-safety applications, and visual-inertial odometry (VIO). VIO is a method for combining visual data with inertial sensors to determine movement and paths more efficiently. More features will be available soon, and registered developers with Perception Studio will get early access.

The Perception Studio Beta Version is available for download to registered developers and includes downloadable binary files.

 

AI-Powered Perception

RealSense expects these vision products to be widely adopted in industrial robots, humanoids, manipulators, and AMRs due to their high-speed, high-resolution edge-detection capabilities. Quality assurance and control departments will appreciate the increased throughput and robust sensing across challenging environments enabled by the flexibility in standoff distance and the IP65 case rating.