Home / Products / AI Video Analysis Algorithm Box

Products

AI Video Analysis Algorithm Box

The AI edge computing server box (AIbox-03) is an intelligent application server integrating 100+ AI algorithms. With an 8-core 2.4G processor and up to 6T of compute, it deploys AI to the edge quickly, plug and play, for fast AI enablement.

2026-10-06 · Zhcloudlink

AI Box / May 2024

Model: AIBox-03

The AI edge computing server box (AIbox-03) is an intelligent application server integrating 100+ AI algorithms. With an 8-core 2.4G processor and up to 6T of compute, it deploys AI to the edge quickly, plug and play, for fast AI enablement.

  • 6T compute
  • Octa-core 2.4GHz
  • 100+ algorithms
  • Northbound streaming
AIBox-03 AI video analysis algorithm box

2. Product Features

  • Supports multiple pretrained AI computer-vision algorithm models.
  • Supports ingest over RTSP, WebRTC, RTMP, HTTP-FLV, Websocket-FLV, HLS, and similar protocols.
  • Compute and application in one: compute and the application platform combined, for integrated edge deployment.
  • Supports remote device upgrade (when the device is online)
  • Open interfaces, supporting secondary development and integration
  • Training and inference in one: supports custom AI algorithms. (A training platform can be provided, with self-service annotation and self-service training.)
  • Storage and compute in one: supports multi-channel video stream aggregation (NVR), and video storage and playback. (An SD card is required.)
  • Different cameras can be assigned different algorithm analysis tasks
  • Supports northbound streaming (RTSP/GB28181)

3. Specifications

ProductAI edge computing box
System parameters
SystemARM Linux (Ubuntu 20.2, Debian 10)
CPUOcta-core, frequency up to 2.4GHz
GPUMali-G610 MC4
NPU6 Tops AI compute, triple-core architecture
StorageStandard 8GB LPDDR4, 128GB storage (eMMC)
Display interface1 HDMI 2.1, up to 8K output
Interfaces4 USB2.0, 2 USB3.0, 1 USB OTG
2 RS232 serial ports (RS485 optional)
Network
Network2 RJ45 + WIFI, optional 4G all-network

4. Product Photos

235mm*150mm*62.5mm (L*W*H
235mm*150mm*62.5mm (L*W*H
Interface diagram
Interface diagram

5. AI Algorithms

1. Inference capability

Multiple AI algorithms, with customization for the application scenario.

Standard algorithm setPeople: crowd gathering detection, illegal intrusion detection, person climbing, absence-from-post detection, monitoring-room absence detection, elevator entrapment detection, entry and exit detection, sleeping-on-duty recognition, smoking recognition, and other behavior analysis
Vehicles: muck truck recognition, e-bike entering an elevator recognition, vehicle wrong-way recognition, non-motor vehicle illegal parking detection, vehicle illegal parking recognition, jaywalking recognition, heavy machinery recognition, pedestrian and vehicle separation recognition
Objects: exposed earth pile detection, work uniform recognition, reflective vest recognition, face mask recognition, e-bike helmet recognition, insulating glove recognition, safety helmet recognition, fire extinguisher placement detection, protective clothing recognition, cat and dog recognition
Environment: road litter recognition, flame recognition, flame detection, overflowing garbage recognition, corridor clutter recognition, open-door recognition for equipment rooms / power rooms / unit entrances, street-vending recognition, smoke recognition, smoke detection

2. Intelligent Computing Platform

Algorithm configuration
Algorithm configuration
Video aggregation
Video aggregation
Video analysis task configuration
Video analysis task configuration
Alert push
Alert push

3. Management Platform

Cloud management platform, including algorithm management, algorithm update delivery, and remote box maintenance.

Video storage and playback
Video storage and playback
Algorithm training platform
Algorithm training platform
Product image

6. Algorithm Customization

Custom development process for deep-learning AI vision algorithms:

1. Annotate images and videos sampled on site;

Obtain the dataset, then design the model, train, and so on
Obtain the dataset, then design the model, train, and so on
Product image

After the model is generated, convert it to an RKNN model for the RKNPU chip;

Product image

7. Application Scenarios

Smart community / park

Algorithms and product applications: parking in a fire lane, area intrusion recognition, fall recognition, smoke and open flame recognition, boundary crossing detection, street vending, people counting, elevator e-bike detection, and other algorithms. Widely used for identity verification of people and vehicles at entry and exit, illegal parking, environmental sanitation checks, recognition of violations and uncivil behavior, and fire-hazard detection.

Parking in a fire lane
Parking in a fire lane
Area intrusion recognition
Area intrusion recognition
Fall recognition
Fall recognition
Smoke and open flame recognition
Smoke and open flame recognition
Boundary crossing detection
Boundary crossing detection
Street vending
Street vending
People counting
People counting
Elevator e-bike detection
Elevator e-bike detection

Algorithms and product applications: fire and smoke recognition, hazardous area recognition, loitering recognition, person detection, person lingering, illegal fishing / stranger intrusion, fall recognition, license plate recognition, and other algorithms. Widely used in villas, courtyards, fish ponds, farms, planting and breeding bases, and unattended stores.

Fire and smoke recognition
Fire and smoke recognition
Hazardous area recognition
Hazardous area recognition
Loitering recognition
Loitering recognition
Person detection
Person detection
Person lingering
Person lingering
Illegal fishing / stranger intrusion
Illegal fishing / stranger intrusion
Fall recognition
Fall recognition

Food-service stores

Algorithms and product applications: customer-flow counting, VIP customer recognition, smoking recognition, phone-call recognition, rodent recognition, chef uniform and hat detection, unattended open flame, staff procedure compliance, smoke and open flame recognition, overflowing trash bins, and other algorithms. Widely used in schools, food service, hospitals, and similar sites.

Customer traffic counting
Customer traffic counting
VIP customer recognition
VIP customer recognition
Smoking recognition
Smoking recognition
Phone-call recognition
Phone-call recognition
Rodent recognition
Rodent recognition
Unattended open flame
Unattended open flame
Staff procedure compliance
Staff procedure compliance
Smoke and open flame recognition
Smoke and open flame recognition

Smart campus

Student behavior analysis

Jumping detection, running recognition, fall recognition, running recognition, climbing recognition, fighting recognition, fire and smoke detection, smoking recognition, mobile phone recognition, tripwire detection, area detection;

Classroom behavior analysis

Raising a hand, standing up, head on the desk, reading, writing, looking up, looking down, using a phone

Teacher behavior analysis

Board writing, stepping on and off the podium, using a phone, making a phone call, whether seated, presenting PPT

Product image
Product image
Classroom behavior analysis
Classroom behavior analysis

Questions

What is the compute of AIBox03?

The CPU is octa-core, with a maximum frequency of 2.4GHz. The GPU is Mali-G610 MC4. The NPU is 6 Tops, with a triple-core architecture. Standard configuration is 8GB LPDDR4 memory and 128GB eMMC storage.

What operating system does AIBox03 run?

The system is ARM Linux, with Ubuntu 20.2 or Debian 10 optional. Display output is 1 HDMI 2.1, supporting up to 8K.

How does AIBox03 ingest video and connect to the network?

It supports RTSP, WebRTC, RTMP, HTTP-FLV, Websocket-FLV, and HLS ingest, and can assign different algorithms to different cameras. Networking is 2 RJ45 ports plus Wi-Fi, with optional 4G, and it also supports northbound RTSP and GB28181 streaming.

Which scenarios is AIBox03 suited for?

Application scenarios on the page include smart communities and parks, food-service stores, and smart campuses. Algorithms cover people, vehicles, objects, and environment detection, and algorithms can also be customized from on-site samples.