Edge · Application

DeepStream + YOLO Vision Linkage

YOLO detects people in ROI · events to HA · Mesh / Wi-Fi IoT linkage

How it works

Edge Vision AI

What it sees

Camera RTSP feeds enter the edge; focus on people and targets inside configured ROI / Zone.

How it judges

NVIDIA DeepStream real-time pipeline runs YOLO inference — occupied / empty / count and confidence.

How it acts

Detection events go to Home Assistant to drive Mesh lighting and Wi-Fi devices.

ROI detection preview

YOLO ROI
RTSP LIVEIn zone · 2
ROI / Zone
person 0.94
person 0.87

Edge inference pipeline

1RTSP camera

IP camera stream into the edge

2DeepStream

NVIDIA GPU-accelerated video pipeline

3YOLO detection

ROI people / object judgment

4Home Assistant

Local automation hub receives events

5IoT linkage

Mesh / Wi-Fi device actions

Three capability pillars

MXCHIP vertical stack: in-house vision + edge hub + own IoT — project-ready customization and deep secondary development.

ASense

In-house vision & customization

DeepStream + YOLO landed and tuned by MXCHIP edge team; ROI / models / scene rules can be customized per project.

BHub

Edge hub integration & customization

Events enter Home Assistant locally; sensor mapping, automations and scenes are delivered by MXCHIP with on-site secondary development.

CAct

Own IoT vertical loop

HA connects MXCHIP Mesh lighting and heterogeneous Wi-Fi products end-to-end — lighting / alert / energy scenes.

AI video analytics & use cases

YOLO
RTSP camera→
DeepStream + YOLO→
AI results→
Home Assistant→
IoT linkage
目标检测

Object detection

Detect people in a clean office and count occupancy in real time.

Outputs occupied / empty / count and confidence; aligned with Ultralytics Region / TrackZone. Events enter HA as local sensors for lighting, curtains, HVAC and more.

车辆识别

Vehicle recognition

Detect vehicles in villa parking / driveway spots

DeepStream multi-stream RTSP inference; YOLO separates enter / leave / stay. Results drive locks, garage lights and barriers via HA (per project).

行为分析

Behavior analysis

Detect smoking and similar violations in no-smoking corridors

Person detection plus pose / smoke cues; conclusions stay on-edge and enter HA automations via AI results — no always-on public cloud.

告警通知

Alert & notify

Turn vision events into actionable local alerts

Map detection / behavior to HA binary sensors or the event bus — phone notify, sound/light, locks and Mesh actions: see → judge → act.

Connect vision events to local automation

Edge DeepStream + YOLO detects people in ROI; HA links IoT — book an advisor for the deployment path.

Book an advisor

Connect 10 billion devices for smart life

Serve 100 enterprises for digital capability

CONTACT US