Multi Object Detection Reference Design

Edge AI Accelerated Object Detection On FPGA

This Multi Object Detection (MOD) reference design enables real-time detection, classification, and tracking of multiple objects in images and video streams at the edge. Built on the Lattice CertusPro-NX FPGA System-on-Module (SOM), the design integrates compute, memory, connectivity, and AI acceleration for robust, low-power vision intelligence in industrial, robotics, automotive, and smart camera applications.

The MOD reference design supports:

  • Simultaneous detection and tracking of multiple object classes
  • Dynamic overlay of bounding boxes and labels on video output
  • Scalable architecture for high-resolution image processing
  • Seamless integration with host processors (e.g., Raspberry Pi CM5) via MIPI and I2C
  • Flexible deployment for counting, classification, and defect detection use cases

Features

  • Identify and track several objects at once, even in dynamic or crowded scenes
  • Efficient AI models optimized for edge devices, enabling always-on operation
  • Identify and track several objects at once, even in dynamic or crowded scenes
  • Supports a wide range of object classes and deployment scenarios
  • Easily connects to existing infrastructure and legacy systems
  • Accurate detection and tracking for safety-critical and monitoring applications

Block Diagram

Multi Object Detection Reference Design