Back to Blog

Nvidia/Sophon + FPGA + AI High-Performance Edge Computing Box: Smart Services for Healthcare

#人工智能#边缘计算

As a leader in domestic sports medicine, dedicated to providing comprehensive clinical solutions for sports medicine, the company is located in the Beijing Economic-Technological Development Area. Over 300 specifications of sports medicine equipment, implants, and surgical instruments for shoulder, knee, foot/ankle, hip, elbow, and hand/wrist joints have received NMPA approval and are clinically applied in thousands of hospitals across 31 provinces, municipalities, and autonomous regions in China. The company has made breakthroughs in several major key technologies, including injection molding, weaving, preparation and processing of absorbable materials, and active device development for sports medicine products. It has established a comprehensive quality management system, with 13 product specifications certified by EU CE, ISO9001, ISO13485, and national GMP, becoming a domestic enterprise capable of providing comprehensive clinical solutions for sports medicine. The company has formed an interdisciplinary composite team centered on design, R&D, production, medicine, clinical application, and marketing. It collaborates on projects with numerous doctors and experts both domestically and internationally, holds over 100 core invention patents, undertakes several national major projects and Beijing Municipal Science & Technology Commission R&D programs, and has more than ten products certified as new technologies and new products by the Beijing Municipal Science & Technology Commission.

Modern healthcare systems, hospitals, and service providers are deploying new tools and building exciting new care models to better serve patients. These strategies focus on Clinical Decision Support (CDS), helping to provide timely, filtered, patient-specific information for clinicians to enhance care. In recent years, this pursuit has seen an increasing number of medical devices introduced into healthcare environments. These devices include tablets, wearables, health monitors, and Artificial Intelligence (AI) imaging systems. Wearable devices can keep clinicians informed of vital patient signs, such as heart rate and blood pressure, alerting healthcare providers before problems arise. Health monitors can collect patient data and trigger corresponding actions based on the results, thereby assisting with remote care—for example, monitoring blood glucose levels and sending information to a companion device (such as an insulin pump) to administer insulin.

AI imaging models can detect potential issues in X-rays and prioritize these images for radiologists or internists to review. Consequently, healthcare systems and health service providers must decide how to manage and fully leverage this unprecedented volume of data. Considering bandwidth overhead, access rights, and privacy, decisions must be made on which data to send to the cloud and which to manage locally. Edge computing brings data processing, analysis, and storage closer to the source where data is generated—for example, a hospital's local servers or a patient's mobile device at home. Edge computing, as a complement to cloud computing, allows IT decision-makers to choose the appropriate computing location for workloads. This strategy can help healthcare systems optimize data collection, storage, and analysis—for a typical hospital, the annual data volume already reaches 50 PB.

    1. Product Overview

The XM-AIBOX-16 Smart Edge Analysis All-in-One Machine is a high-performance, low-power edge computing product. Equipped with the domestic TPU chip BM1684, it boasts INT8 computing power up to 17.6 TOPS, can simultaneously process 16 channels of high-definition video, and supports hardware decoding of 32 channels of 1080P HD video and 2 channels of encoding.

This product highly integrates high-precision AI intelligent algorithms based on computer vision and deep learning networks, as well as an intelligent video comprehensive management platform. The AI intelligent algorithms cover various scenarios such as parks, communities, construction sites, and campuses, and can be combined and configured on demand according to the scenario. The intelligent video comprehensive management platform supports front-end device management, real-time video preview, alarm push, forensic snapshot, online algorithm loading and optimization, data situational analysis large-screen display, etc. The device is easy to operate, plug-and-play, and features rich northbound API interfaces, enabling integration with upper-layer business application platforms.

    1. Product Features

Ultra-High Performance Computing and Codec Capabilities

  • Supports up to 17.6 TOPS INT8 peak computing power or 2.2 TFLOPS FP32 high-precision computing power;
  • Supports full-process handling of up to 16 channels of 1080P HD video;
  • Supports hardware decoding of up to 32 channels of H.264/H.265 1080P@25FPS video;
  • Supports hardware encoding of up to 2 channels of H.264/H.265 1080P@25FPS video.

Rich Built-in AI Algorithms

  • Built-in with over 30 AI algorithms, supporting free matching and custom combinations;

(Supports person structuring / facial recognition / vehicle structuring / license plate recognition / flame detection / smoke detection / smoking detection / phone call detection / mobile phone play detection / unmasked detection / personnel absenteeism detection / personnel sleeping on duty detection / personnel fall detection / personnel electrostatic discharge / area crowd counting / insufficient area crowd / excessive area crowd / abnormal area crowd / area intrusion detection / work uniform detection / safety helmet detection / reflective vest detection / electric vehicle detection / regulated parking (illegal vehicle parking) / entrance/exit passenger flow statistics / perimeter breach intrusion / personnel boundary crossing detection / area loitering detection / fire lane occupation / fire escape route occupation / un-disposed trash detection / overflowing trash bin detection / trash disposal reminder / abnormal camera displacement detection, and other algorithms)

  • Each video channel supports up to 3 AI analysis tasks running simultaneously;
  • Supports up to 16 video AI analysis tasks running simultaneously; if exceeding 16 AI analysis tasks, analysis can be performed by polling.

Rich Interfaces, Flexible Deployment

  • Supports rich interfaces: 1000M Ethernet port, USB3.0/USB2.0, HDMI, RS-485, RS-232;
  • Supports wide operating temperature environment from -20℃ to +60℃;
  • Supports IP30 protection level, supports fanless cooling (subject to specific model);
  • Adapts to support SATA storage, supports 2TB storage capacity (subject to specific model);
  • Optional support for LTE wireless backhaul function (subject to specific model);
  • Northbound interfaces: Supports HTTP protocol, MQTT protocol, GB28281
  • Southbound interfaces: Supports GB28281, Onvif, RTSP

High Reliability, Encryption Protection

  • Supports high-capacity eMMC with developable primary/secondary partition support;
  • Supports abnormal fault alarm and protection handling mechanisms;
  • Supports programmable encryption chip for privacy information protection.

Easy-to-Use Toolchain, Flexible Development

  • One-stop deep learning development toolkit Sophon SDK;
  • Supports mainstream deep learning frameworks such as Caffe/DarkNet/TensorFlow/PyTorch/MXNet/ONNX/PaddlePaddle;
  • Supports mainstream network models for classification and detection, supports custom operator development;
  • Supports Docker containerization for rapid deployment of algorithm applications.
    1. Technical Specifications

Specifications

Model Version

XM-AIBOX-16

Technical Specifications

Chip

TPU

BM1684

CPU

8-core A53@2.3GHz

AI Computing Power

INT8

17.6TOPS

FP32

2.2 TFLOPS

Video/Image Codec

Video Decoding Capability

H.264/H.265: 1080P @960fps

Video Decoding Resolution

8K / 4K / 1080P / 720P / D1 / CIF

Video Encoding Capability

H.264/H.265: 1080P @50fps

Video Encoding Resolution

4K / 1080P / 720P / D1 / CIF

Image Encoding/Decoding Capability

480 frames/sec@1080P

Max Image Decoding Resolution

32768 * 32768

Memory and Storage

Memory

12 GB

eMMC

32 GB

External Interfaces

Network Port

10/100/1000Mbps adaptive *2

USB

USB3.0 *2

N/A

Storage

MicroSD *1

Display

HDMI *1

N/A

Serial Port

RS232 *1/RS485 *1/Custom I/O

Extended Storage

SSD (Optional)

2TB (Optional)

N/A

Wireless Functionality

4G/5G Wireless Module (Optional)

Mini-PCIe 4G Module

Antenna (Optional)

SMA Female *2 (Wi-Fi)

SMA Female *4 (5G)

SIM

Standard SIM Card Slot

Wireless Functionality (Non-Optional)

5G Wireless Module

M2.5G Module

Physical Specifications

Dimensions

LengthWidthHeight

188 mm * 148 mm * 44.5 mm

Power Supply and Consumption

Power Supply

DC 12V

Typical Power Consumption

≤23.5W

≤19W

Note: Hard drive and wireless functions are optional and not standard product configurations. Typical power consumption does not include hard drive or wireless module power consumption.