SeerSense® DS80 Module - Xvisio Technology

Multi-sensory fusion defines a new benchmark for spatial perception

DS80 has advantages in depth perception (6-meter field of view), ruggedness (industrial-grade protection, adaptable to harsh environments), compact design, cross-platform support (Windows/Linux/Android), and rich SDK.

Global spatial cognition center
VSLAM 2.0 builds 3D semantic maps in real time

DS80 uses high-speed, low-latency, and high-precision self-developed SLAM technology.
The high-performance SOC has greatly improved computing power and supports multiple advanced working modes such as high-speed and high-precision independent VIO, SLAM/CSLAM, end-cloud collaboration, real-time loop detection, high-speed QR code/spatial anchor assisted positioning, etc.
Deep collaboration among multiple sensors enables faster, more accurate and more stable positioning.

Spatial positioning accuracy test report

I HAVE	Cycles	Ave RTE (%)	Max RTE (%)	Ave ARE (°)	Max ARE (°)
SN: 21004424	1	0.12	0.33	0.51	1.58
	2	0.11	0.48	0.47	2.25
	3	0.12	0.50	0.53	2.18
	Average	0.11	0.44	0.51	2.00
Remark	Test conditions: 10x5m office test environment, 20-30m moving trajectory RTE: Relative Translation Error relative translation error ARE: Absolute Rotation Error

Adaptive environment understanding dual-mode deep learning engine for intelligent decision-making

Passive binocular vision:
With a large field of view (110°), high resolution (640*480@60fps), and an effective distance of 5.5 meters (error <3%), it performs excellently in bright light environments and is suitable for modeling and obstacle avoidance in open scenes.

Active iTOF (Sony VGA):
940nm anti-interference wavelength, ultra-high accuracy (error <1%) within 0.2-4 meters, 30fps frame rate, and output of depth maps, point clouds, and IR images. Its accuracy in low-light and indoor environments far exceeds that of binoculars. It supports switching between single-frequency (1.5m) and dual-frequency (4m) modes.

The dual engines can work simultaneously and seamlessly integrate with VSLAM to provide real-time environmental depth information, laying a solid foundation for navigation, obstacle avoidance, and 3D reconstruction.

Localized AI engine gives "eyes" the ability to think

The independent hardware CNN engine is a key breakthrough! It breaks away from the dependence on the computing power of the main control system and realizes localized real-time AI reasoning. It supports the access of multiple video sources such as fisheye, TOF, RGB, and can deploy multiple AI models in parallel. Combining VSLAM and depth information, it realizes advanced intelligent functions such as 3D gesture recognition and tracking, human skeleton extraction, 3D face anti-counterfeiting recognition, object recognition and spatial semantic segmentation, and scene understanding. Perfectly compatible with OpenCV & Open VINO, which allows for easy deployment of models from open source communities (such as Open Model Zoo) or user-trained models, with plug-and-play functionality, significantly shortening the R&D cycle.

Efficient integration and expansion Out-of-the-box and flexible expansion

Intelligent video compression

Hardware-based H.265 video compression significantly saves valuable transmission bandwidth and reduces the overall system load.
Plug and Play

Localized processing capabilities significantly lower the development and integration threshold.
Enrich SDK

Supports Windows/Linux/Android, and provides wrappers for ROS/ROS2, Unity, C/C++, Python, Java JNI, etc. API, seamlessly connect to various development frameworks.
Highly scalable

The USB Type-C is used to connect to the main controller, and sensors such as lidar can be connected through the expansion interface. It supports multi-machine cascading to meet complex system requirements.
Global Certification

Passed CE, FCC, RoHS, CB, FDA Class I laser safety certifications, and is accepted in the global market.

Empowering thousands of industries and leading the future

SeerSense® DS80 is a dazzling manifestation of Xvisio's spatial perception technology strength. The core technology matrix behind it is not only the "smart eye" for humanoid robots to achieve autonomous navigation, dexterous operation, and intelligent interaction, but also has a wider range of application potential.

Smart Manufacturing

Humanoid robot/AMR navigation, industrial inspection, and automated sorting.
Daily life

Smart access control and personnel behavior analysis.
Medical

Surgical navigation, rehabilitation assistance, and telemedicine perception.
Logistics and Retail

Volume measurement, shelf analysis, and unmanned delivery.
AR/VR & the Metaverse

High-precision positioning tracking and immersive interaction.
Office and entertainment

Virtual shooting, somatosensory games, virtual meetings.
Smart Home

Service robots and intelligent monitoring.

Product composition

SeerSense® DS80
SDK
Tool Chain

System specifications

Features	Specification	Remark
processor	High performance SOC
camera
-Binocular fisheye camera	1280x800/640x400, 60fps DFOV150°/HFOV130°/VFOV74°	Monochrome, global exposure 80mm baseline
-Color camera	8MP（4192x3104）, 30fps DFOV79.9°/HFOV68°/VFOV53°	Camera: 8.2MP/13MP (MAX) video: VGA/720P/1080P @ 30fps
-Depth camera	640x480/320x240, 30fps DFOV78°/HFOV64°/VFOV50°	depth error=<1% 940nm
IMU	9 Axis	1000 Hz for SLAM fusion
CNN Engine	Dual engine	Parallel processing
Binocular Depth Engine	VGA/720P SGBM（Semi-Global Block Match）	50/60fps, 6.5 meter
SLAM Engine	500-1000Hz millimeter-level accuracy	Local processing, multiple working modes
Power consumption	Avg 2.8W	Full load
powered by	5V 3A
interface	USB Type C	Camera data stream, depth, AI image, 6DoF output
interface	Debug Port	UART, Sync Pin
Weight and size	93g \| 119Lx23Dx24H	LxDxH（mm）
SDK	VIO, CSLAM, 3D reconstruction, depth/point cloud (SGBM/TOF), object recognition, gesture, plane detection, RGBD
Toolchain	Open VINO, Open CV, Xvxdk-viewer Demo Tool