Vision AI Operations — YOLOv8 Skeleton Pose

Edge neural inference · 17-point worker pose tracking & fall detection

AI ENGINE OFFLINE
Start Local YOLOv8 Pose Engine

Run the Python pose service in your terminal to enable hardware-accelerated Apple Silicon GPU inference:

npm run ai

Vision & Pose AI

Direct ESP32-CAM MJPEG stream

SOURCE:
Live Feed
ESP32-CAM RAW
Personnel Safety MonitorZONE CLEAR
0
STANDING
0
CRAWLING
0
FALLEN / DOWN
17-Point Skeleton Map
COCO POSE
Head / Cranium
Nose, Eyes, Ears (0-4)
Torso & Spine
Shoulders, Hips (5,6, 11,12)
Upper Limbs
Elbows, Wrists (7,8, 9,10)
Lower Limbs
Knees, Ankles (13,14, 15,16)
AI Architecture
v8.4.143
Neural ModelYOLOv8 Nano Pose (yolov8n-pose.pt)
AccelerationApple Silicon Metal (MPS GPU)
Input Resolution640 × 480 px
Inference Latency0 ms (~0 FPS)
Stream ProtocolMJPEG over HTTP (/api/ai/stream)
Connection: Connected
Env (DHT22): Online
Gas (MQ-135): Online
IMU (MPU-6050): Online
Camera: Streaming
System: Normal