Face Tracking

Lattice sensAI Demo

ECP5 based Convolutional Neural Network (CNN) acceleration for face tracking with 8 convolution layers implemented in 8 Neural Network (NN) engines on ECP5-85K FPGA.

Standalone operation based on Embedded Vision Development Kit tracking face movements.

Green box indicates location of face in image.

Features

  • Runs @ 14fps with 90 x 90 RGB Input
  • Configuration files provided for rapid implementation on Embedded Vision Development Kit
  • Up to 200fps possible with 32 x 32 RGB Input
  • Total ECP5 power consumption of 0.85 W
Lattice sensAI

Jump to

Video

Face Tracking Using ECP5 and CNNs

  • This demonstration identifies and tracks a human face
  • The inferencing is done using Convolutional Neural Networks implemented in the Embedded Vision Development Kit’s ECP5 FPGA
  • Power consumption is less than 1W

Block Diagram

Face Tracking Block Diagram

Documentation

Quick Reference
Downloads
TITLENUMBERVERSIONDATEFORMATSIZE
Select All
EVDK Based Face Tracking Demonstration User Guide
FPGA-UG-020431.39/30/2018PDF1.7 MB

*By clicking on the "Notify Me of Changes" button, you agree to receive notifications on changes to the document(s) you selected.

TITLENUMBERVERSIONDATEFORMATSIZE
Select All
EVDK Based Face Tracking Demonstration Bitstreams
1.29/26/2018ZIP12.4 MB