KORTHOS / ECOSYSTEM / PRODUCT

ZED 2i

About

About ZED 2i

ZED 2i uses two synchronised cameras and a wide stereo baseline to calculate dense depth from nearby objects to distances of roughly 20 metres. Its onboard inertial sensors combine with visual odometry for six-degree-of-freedom tracking and spatial mapping as the host robot or drone moves.

The camera is available with two lens fields of view and an IP66-rated enclosure for outdoor or industrial installations. The ZED SDK provides depth, positional tracking, object detection and mapping on NVIDIA and other supported compute platforms, allowing one sensor to support depth and perception workflows.

Capabilities

Source-backed capabilities

  • Capability areas: Robot perception, Depth Sensing, Stereo Vision
  • Workflows: Mobile Robot Navigation
  • Operating environments: Robotics, automation and edge-AI systems that need sensing, depth perception, vision processing or local compute.
  • Target markets: Robot builders, machine-vision teams, automation engineers and edge-AI developers.
  • Deployment model: Commercial or available product positioned for customer use, integration or developer adoption.
  • Object manipulation

    Research demonstration - 2025-09-02 - The paper studies geometry-aware perception for robotic manipulation beyond conventional 2D colour observations.

    Source
  • Whole-body loco-manipulation

    Research demonstration - 2025-06-29 - The work connects ZED 2i perception to learned whole-body manipulation policies.

    Source
  • Object manipulation

    Research demonstration - 2025-06-29 - ParticleFormer presents a 3D point-cloud world model for multi-object, multi-material robotic manipulation.

    Source
  • Robotic grasping

    Research demonstration - 2023-07-21 - The study compares two machine-learning and two analytical vision-based robotic grasping algorithms under a common benchmarking protocol.

    Source

Observed use

Exact-product configurations are shown without inferring supply, purchase, customer status, standard fit, deployment or partnership.

  • Autonomous greenhouse-sprayer prototype

    Prototype - Mounted - Physical - Maruyama's greenhouse-sprayer prototype used a rear-mounted Stereolabs ZED 2i with a 2.1 mm lens for perception and localisation.

    Source

Research

  • Comparing Commercial Depth Sensor Accuracy for Medical Applications

    2026-06-11 - Research Publication - Depth estimation has numerous medical and surgical applications. We benchmark four depth sensors on a porcine bone specimen, a porcine belly specimen, and a silicone kidney phantom using stylus-sampled references.

    Source
  • Finite-time Stable Pose Estimation on TSE(3) using Point Cloud and Velocity Sensors

    2026-02-10 - Research Publication - This work presents a finite-time stable pose estimator (FTS-PE) for rigid bodies undergoing rotational and translational motion in three dimensions, using measurements from onboard sensors that provide position vectors to inertially-fixed points and body velocities. The FTS-PE is a full-state observer for the pose (position and orientation) and velocities and is obtained through a Lyapunov analysis that shows its stability in finite time and its robustness to bounded measurement noise.

    Source
  • Reliability assessment of markerless technologies in biomechanical motion analysis: a performance comparison

    2026-01-12 - Research Publication - Introduction: Body tracking systems are utilized in various areas such as health and sports. In these areas, it is crucial to identify the most accurate and appropriate system for the intended design of the study. Methods: This study evaluates the reliability of markerless motion capture systems compared to a marker-based system (OptiTrack) as the gold standard. The systems assessed include StereoLabs ZED 2i, MediaPipe (2D and 3D), and YOLOv8 Pose (8n and 8x-p6). Simple upper and lower limb movements were analyzed under controlled conditions in a single participant. Intraclass Correlation Coefficient (ICC) values and Bland-Altman analysis were mainly used to evaluate agreement between systems, with some other techniques explored to provide additional perspective. Results: The results indicate that StereoLabs ZED 2i achieved the highest reliability among markerless systems (ICC: 0.92 for

    Source
  • A Modular Architecture Design for Autonomous Driving Racing in Controlled Environments

    2025-12-03 - Research Publication - This paper presents a modular autonomous driving architecture for Formula Student Driverless competition vehicles operating in closed-circuit environments. The perception module employs YOLOv11 for real-time traffic cone detection, achieving 0.

    Source
  • Think Fast: Real-Time Kinodynamic Belief-Space Planning for Projectile Interception

    2025-11-30 - Research Publication - Intercepting fast moving objects, by its very nature, is challenging because of its tight time constraints. This problem becomes further complicated in the presence of sensor noise because noisy sensors provide, at best, incomplete information, which results in a distribution over target states to be intercepted.

    Source
  • High-Precision 3D Recognition of Road Potholes Based on Binocular Vision and Cross-Modal Feature Fusion

    2025-11-26 - Research Publication - Abstract. Efficient detection and accurate three-dimensional characterization of road potholes are crucial for road maintenance and traffic safety. To address the issues of high cost and poor environmental adaptability in existing detection methods, this study proposes a lightweight pothole detection and 3D reconstruction method based on binocular stereo vision and deep learning. A ZED 2i binocular camera was used to build a vehicle-mounted acquisition system, combined with the Mask R-CNN model to achieve pothole detection and pixel-level segmentation. The 3D point cloud of potholes was reconstructed using the principles of binocular stereo vision, and a dynamic mesh density method was proposed to optimize surface area calculation. Additionally, the RANSAC algorithm was employed to fit the ground plane and extract depth parameters. Experimental results demonstrate that this method can ac

    Source
  • Imagination at Inference: Synthesizing In-Hand Views for Robust Visuomotor Policy Inference

    2025-09-19 - Research Publication - Visual observations from different viewpoints can significantly influence the performance of visuomotor policies in robotic manipulation. Among these, egocentric (in-hand) views often provide crucial information for precise control.

    Source
  • A Real-Time Diminished Reality Approach to Privacy in MR Collaboration

    2025-08-21 - Research Publication - Diminished reality (DR) refers to the digital removal of real-world objects by compositing background content in their place. This thesis presents a real-time, inpainting-based DR system designed to enable privacy control in shared-space mixed reality (MR) meetings.

    Source
  • ParticleFormer: A 3D Point Cloud World Model for Multi-Object, Multi-Material Robotic Manipulation

    2025-06-29 - Research Publication - 3D world models (i. e.

    Source
  • A stereo-vision SLAM method based on Manhattan-plane constraints and point-plane collaborative optimization

    2025-05-06 - Research Publication - Abstract Previous research on visual simultaneous localization and mapping (SLAM) based on Manhattan framework constraints relied on the Manhattan world assumption, which only constrained the rotation matrix. However, although this can reduce drift in the rotation matrix, the translation matrix inevitably accumulates errors. To make full use of the structural information in the scene, this study proposes a SLAM method that utilizes two or three orthogonal plane features extracted from coplanar line features in stereo images to construct and match Manhattan plane (MP) features. By using three orthogonal MP features, the method can decouple both the absolute rotation and translation matrices simultaneously in a structure-from-motion manner, referred to as MHT-SP (MP constraint based on stereo-plane). To address the low tracking and matching rate of plane features in large-scale scenes, a m

    Source
  • Low-Cost 3D Depth Sensors for Mobile Applications and Control Systems – Accuracy Assessments Using Surveying Techniques

    2025-02-11 - Research Publication - This article focuses on low-cost LiDAR (light detection and ranging) sensors and 3D depth cameras. Particular attention was paid to their accuracy and compliance with the technical specifications that were provided by their respective manufacturers. The following devices were tested: Stereolabs ZED 2i, Stereolabs ZED, and Intel RealSense D435i depth cameras, and the Intel RealSense L515 LiDAR sensor. An experiment was carried out to measure a geometrically diverse environment (which is typical for in-motion imaging) where both the measurement range and the distortion that is generated by each device’s algorithms on edges, folds, planes, and 3D objects could be evaluated. Depth sensors are often used with excessive confidence as to their geometric reliability. The aim of this work is to assess the actual accuracy of such sensors, which may constitute the ground truth for accuracy losses t

    Source
  • A Benchmarking Study of Vision-Based Robotic Grasping Algorithms: A Comparative Analysis

    2023-07-21 - Research Publication - We present a benchmarking study of vision-based robotic grasping algorithms and provide a comparative analysis. In particular, we compare two machine-learning-based and two analytical algorithms using an existing benchmarking protocol from the literature and determine the algorithms strengths and weaknesses under different experimental conditions.

    Source
  • ZED 2i paper uses depth sensing for robotic EV charging-socket detection

    2023-02-10 - Research Publication - This paper covers ZED 2i paper uses depth sensing for robotic EV charging-socket detection for ZED 2i in Robot learning.

    Source
  • ZED 2i skeleton-tracking paper evaluates automatic body-orientation detection indoors

    2022-05-17 - Research Publication - This paper covers ZED 2i skeleton-tracking paper evaluates automatic body-orientation detection indoors for ZED 2i in Robot learning.

    Source

Context

Market context

Flagship stereo camera for robotics and industrial environments.

  • Target environment: Robotics, automation and edge-AI systems that need sensing, depth perception, vision processing or local compute.
  • Workflow context: Depth sensing, visual perception, edge processing, object detection, tracking and robot awareness.
  • Customer context: Robot builders, machine-vision teams, automation engineers and edge-AI developers.
  • Deployment model: Commercial or available product positioned for customer use, integration or developer adoption.
  • Commercial maturity: Publicly presented as commercial or available, with maturity depending on deployment scale and support coverage.
  • Adoption constraints: Adoption depends on accuracy, latency, environmental robustness, SDK quality, compute budget and integration with the robot stack.
  • Market position: ZED 2i sits in the perception and compute layer that helps robots understand or process the physical world.
  • Adjacent products: ZED cameras, Ouster lidar, Cognex VisionPro, NVIDIA Jetson

Comparable Systems

Related Systems