Run RTSP Camera Capture and AI Inference

This guide explains how to set up and run the RTSP camera capture and AI inference pipeline.

Download AI Model and Resources

First, download the necessary AI models and supporting libraries.

git clone https://github.com/open-edge-platform/edge-ai-libraries.git
cd edge-ai-libraries/microservices/dlstreamer-pipeline-server
wget -c https://github.com/open-edge-platform/edge-ai-resources/raw/a7c9522f5f936c47de8922046db7d7add13f93a0/models/INT8/pallet_defect_detection.zip
unzip -q pallet_defect_detection.zip -d models/
cd docker

Configure the Environment

Before running the services, you need to configure the environment variables.

Update .env file

Set the MQTT_PORT in the .env file. If you are behind a proxy, configure the proxy settings as well.

MQTT_PORT=1883
# http_proxy=...
# https_proxy=...

Copy Configuration Files

Copy the ptp_frame_timestamp.py and config.json from the deterministic-threat-detection module to the current docker directory.

cp edge-ai-suites/metro-ai-suite/deterministic-threat-detection/rtsp_camera_stream/ptp_frame_timestamp.py .
cp edge-ai-suites/metro-ai-suite/deterministic-threat-detection/rtsp_camera_stream/config.json .

Note: Make sure to update the RTSP camera <rtsp-camera-username>, <rtsp-camera-password>, and <rtsp-camera-url> in the config.json file before proceeding. Also, add the RTSP camera IP to the no_proxy environment variable if you are behind a proxy.

Update Docker Compose File

Comment the existing resources folder mapping and add volume mappings to the docker-compose.yml file to make the custom script and configuration available to the dlstreamer-pipeline-server container.

services:
  dlstreamer-pipeline-server:
    # ... existing configuration ...
    volumes:
      # - "../resources:/home/pipeline-server/resources/"
      - "../models:/home/pipeline-server/resources/"
      - "./ptp_frame_timestamp.py:/home/pipeline-server/ptp_frame_timestamp.py"
      - "./config.json:/home/pipeline-server/config.json"
      # ... other volumes ...

Run the Services

Start the services using Docker Compose.

docker compose up -d

Start the RTSP Camera Pipeline

Finally, start the pipeline by sending a POST request to the pipeline server.

curl -k http://localhost:8080/pipelines/user_defined_pipelines/rtsp_camera_pipeline -X POST -H 'Content-Type: application/json' -d '{
    "destination": {
        "metadata": {
            "type": "mqtt",
            "publish_frame": true,
            "topic": "tsn_demo/camera/inference"
        }
    },
    "parameters": {
        "detection-properties": {
            "model": "/home/pipeline-server/resources/deployment/Detection/model/model.xml",
            "device": "CPU"
        }
    }
}'

Note: Update the topic name if you are running the pipeline on multiple machines.