Skip to content

Raivin Quick Start

Raivin
Raivin

The Raivin configuration further extends the perception capabilities of the Maivin platform through the addition of the integrated radar module or LiDAR module. The EdgeFirst Perception Middleware is augmented for Raivin configurations with the fusion service which projects the radar and LiDAR point clouds onto the vision model output to classify targets, produce 3D bounding boxes, and an occupancy grid. The RadarExp Fusion Model additionally provides low-level radar data fusion with the vision data. The low-level radar data is represented as the range and doppler data cube, the RadarExp model fuses this data with the vision data to provide a more robust perception stack. The traditional point-cloud data from the radar is also available for use by the Raivin perception stack, and is especially useful for augmenting the object detection and tracking capabilities through the additional parameters offered by the point cloud data.

System Architecture

The Raivin runs a Linux operating system based on Toradex's Torizon OS 7, we refer to the operating system as Torizon for Maivin. The operating system packages include common libraries and tools you would expect to find on a typical embedded Linux operating system focused on computer vision and AI (Python, OpenCV, GStreamer, etc...) along with the Docker container engine for user applications. Refer to the Release Notes for the versions included in the current release and to Software Updates for keeping the platform up to date.

Central to the EdgeFirst Platforms is the EdgeFirst Perception Middleware, what we call our collection of applications and libraries used in the implementation of the perception stack. On the Raivin these applications run natively as systemd services, the radar or LiDAR publisher and the fusion service are enabled during provisioning in addition to the base Maivin services. The details of the low-level libraries are covered in the EdgeFirst Perception Developer Guide, for this document we focus on describing the application services and how they fit together to deliver the perception middleware.

In this Quick Start, you will find instructions for:

  1. Setting up your Raivin
  2. Visiting the Raivin's Web UI
  3. Copying a Sample Dataset in EdgeFirst Studio for Training
  4. Training Fusion Model
  5. Validating Fusion Model
  6. Deploying Fusion Model