Docker on O3R

• Getting started with Docker
  • Build and run a Docker container for the O3R platform
    • A basic container
      • Build the container
        • Troubleshooting: proxies
      • Run a container
      • Save a container
    • Load and start a container
    • Add features to the container
    • Install ifm3d in the container
    • Building on top of the ifm base image
• Deployment on board
  • Using SCP
  • Using a Docker registry
    • Configuration
    • Create a local Docker registry
• Docker logging configuration
  • Docker default logging
    • Docker logging drivers
    • Docker logging configuration
    • Run a Docker container with logging driver configuration
    • Double check the Docker logging configuration for a container
• Autostarting the container
  • Docker compose
    • Sample docker-compose.yml
  • Start the container
  • Auto start the container after a reboot of the VPU
  • Consistently save data on the VPU within a container
• Resource Management on the VPU
  • Resources availability for user code
    • Available RAM
    • CPU and GPU resources
      • Cameras only, no ifm application
      • ODS application
  • Process pinning
  • Docker quotas
• Enabling GPU usage
  • Using the GPU of the VPU
    • Dockerfile sample
    • Start the container with the NVIDIA runtime
      • Using docker run
      • Use docker-compose to specify the runtime
• TensortRT: DL / ML model deployment
  • Building a TensorRT container
    • NVIDIA base containers
      • Compatibility Matrix
      • Verify the functionality
  • Using TensorRT in a container on the VPU
    • Adaptations for the OVP8xx architecture
    • Runtime inference cycle times
  • Calculating the inference on OVP81x VPU using YOLOv11 ONNX Model file
    • Example runs
    • Inference timings
  • Deepstream-l4t

This section explains the handling and deployment of customer specific containers for the O3R.