หมวดหมู่ วิธีการและสไตล์ วิดีโอ

In this project, we build a complete real-time AI object detection streaming pipeline using IoT and edge-to-server processing.

We use an ESP32-S3 with OV5640 camera module to capture images and send them via MQTT protocol to a Node-RED server.

On the server side, Node-RED processes incoming images using TensorFlow COCO-SSD model for object detection. The system automatically draws bounding boxes and labels on detected objects.

The processed images are then forwarded as a MJPEG stream to Restreamer, which converts and broadcasts the video stream to HLS (HTTP Live Streaming) for web playback.

🔧 System Architecture

ESP32-S3 + OV5640 Camera
→ MQTT Image Publishing
→ Node-RED Server
→ TensorFlow COCO-SSD Inference
→ Image Annotation (Bounding Boxes)
→ MJPEG Stream Output
→ Restreamer
→ HLS Live Stream

🚀 Features

• Real-time image transfer via MQTT
• AI object detection using COCO-SSD
• Live bounding box annotation
• Node-RED automation pipeline
• MJPEG to HLS streaming conversion
• Web-ready live video output

💡 Use Cases

• IoT AI camera systems
• Smart surveillance prototypes
• Edge AI + cloud hybrid systems
• Industrial monitoring demos
• Smart home / robotics projects