AI edge computing unit
边缘 AI · 本地智能

AI edge computing unit

Small form factor, big compute — powering endless IoT intelligence for effortless living.

SEDGE-NJ10-8G Standard

AI edge computing unit

SEDGE-NJ10-8GStandard

Small form factor, big compute — powering endless IoT intelligence for effortless living.

Book a demo

TBD · 评审占位 · 型号命名说明

SEDGE-NJ10-8G Standard 型号评估

✅ 优点:结构紧凑,可读性不错

SEDGE:边缘智能产品线 · NJ10:NJ+10 · 8G:内存 · Standard:标准版

/products/sedge-nj10-unit · SKU sedge-nj10-8g

What it can do

Turn “seeing people” into “devices acting on their own” — for meeting rooms and offices

Understand scenes locally

Count people from camera video on-site — no cloud required to know occupied vs empty.

Switch devices automatically

Results go through the Mesh gateway to lights, HVAC, curtains, sockets and more.

Save energy, stay at ease

Turn lights and HVAC off when empty; video stays on-site for faster, calmer control.

How it works

IP摄像头(RTSP)→ AI边缘计算网关(HA + Local AI)→ 蓝牙 Mesh 网关 → Mesh 子设备

IP摄像头(RTSP)

视频接入 · 会议室 / 办公区画面

→

AI 边缘计算网关

Home Assistant
Ultralytics YOLO
SEDGE-NJ10-8G
↓

蓝牙 Mesh 子设备

灯具
开关
温控

灯具 / 开关 / 温控 · 有人开 · 无人关

←

蓝牙 Mesh 网关

网关
中控屏

庆科智能网关 / 中控屏 · 下发开/关指令

例如:会议室检测到无人 → SEDGE 本地确认 → 经蓝牙 Mesh 网关关闭空调与灯光。

软件应用对齐 HA 本地生态:IP摄像头(RTSP)→ AI 视频分析与应用 → AI边缘计算网关 → 蓝牙 Mesh 网关 → 蓝牙 Mesh 子设备

Ecosystem Access

HA Local Ecosystem Access

Home Assistant 边端中枢 · 本地优先、隐私可控 · Mesh / Wi-Fi / 摄像头一盘棋

LIVE
区内有人 · 2
ROI 区域
person 0.94
person 0.87
YOLO ROI边侧应用

DeepStream + YOLO 视觉联动

摄像头 RTSP 进边端,ROI / Zone 内人员检测后事件交给 Home Assistant,再联动 IoT。完整五步管线与案例见独立应用页。

LEARN MORE

Advantages of on-device AI

People counting runs on the edge unit — faster, calmer, without sending all video to the cloud

Faster response

Recognize on-site occupancy instantly so devices act without waiting on the cloud.

More private video

Video stays on-site; only results like headcount drive linkage — sensitive frames need not go to the cloud.

Works offline

Local model + Mesh gateway keep meeting-room energy logic running through network blips.

Easier cost control

Daily people counting does not bill per cloud-LLM call — steadier long-term budgeting.

On-device model (primary)

Sees scenes, counts people, triggers switches — the workhorse for daily energy saving.

Cloud LLM (optional)

Connect to the cloud only for richer strategy tips — not required to switch lights or HVAC.

In short: make “see it, switch it” solid on-device first; add cloud when needed.

Three applications

Lead with people-count linkage; the other two are equally practical

Primary

People-count linkage

Count people on-device; notify lights/HVAC via SGW2052. Ideal for empty-room energy saving.

App 2

Pre-meeting prep

Before meetings, turn lights and HVAC on by schedule so the room is ready.

App 3

Space utilization

Track which rooms are busy or idle locally to plan desks and space.

技术参考

底层算力平台开发资料可参考 NVIDIA 官方文档(Jetson Nano / 嵌入式系统产品开发)。

NVIDIA · Jetson 产品开发文档 →

Connect 10 billion devices for smart life

Serve 100 enterprises for digital capability

CONTACT US