Human Face Detection

Lattice sensAI Reference Design

Add Human Face detection (or any specific object of interest) to any Device – The human face detection reference design enables systems to always search for the presence of a human face via a CMOS image sensor. With an updated training model using the deep learning framework, Caffe (without any other changes), systems can use our design and detect any object of interest.

Reduce Power without Compromising Performance – Systems can now enable artificial intelligence with an always-on image sensor, while consuming less than 1 mW of active power.

Production-priced, Space-efficient Design – The complete inference engine fits inside of the 2.15 mm x 2.55 mm iCE40 UltraPlus-5K FPGA and can work with various popular image sensors. Featuring proven functionality with Omnivision’s OVM7692 and Himax’s HM01B0.

Features

  • Frame-rate detection can be adjusted and set at 5 frames per second to keep the power consumption under 1 mW
  • Hardware platform used to demonstrate this reference design is the iCE40 UltraPlus MDP board, which uses the Omnivision OVM7692
Lattice sensAI

Jump to

Video

Human Presence Detection Using ECP5 and CNNs

  • This demonstration processes video images and identifies the presence of a human
  • 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 Detect Block Diagram

Documentation

Technical Resources
TITLENUMBERVERSIONDATEFORMATSIZE
Select All
Human Face Detection Using Compact CNN Accelerator IP - Documentation
FPGA-RD-020341.25/20/2019PDF540.2 KB
Human Face Detection Using Compact CNN Accelerator IP Project Files
1.09/25/2018ZIP1.1 MB

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