Hand Gesture Detection

Lattice sensAI Demo

Hand Gesture detection in smart IoT devices – This demo uses artificial intelligence (AI) to implement hand gesture detection algorithm. FPGAs have parallel data processing ability, making them more power efficient at such tasks compared to a microprocessor.

Always-on, local intelligence improves security – Designing AI to the network edge with an iCE40 UltraPlus FPGA can dramatically lower power consumption for always-on operation while reducing response time. Keeping processing local also improves security.

Compact CNN in a tiny FPGA – The Lattice inference engine with VGG like CNN architecture is implemented in a 48-pin QFN package on a low cost UPduino 2.0 board.


  • Accelerated, low-power hand gesture (open/close) detection at the network edge using neural network model
  • Configuration files enable rapid implementation on UPduino 2.0 board with Himax HM01B0 image sensor
  • Input resolution of 32x32x1 connects to a VGG like CNN with 6 convolutions, 4 max pooling and fully connected layers
  • With integrated 128 K bytes of memory, weights/activations can be stored directly inside of iCE40 UltraPlus FPGA
  • Power consumption of 3.3 mW @ 5 frames per second. The reference design can be optimized between power and response time depending on system needs
  • The neural network can be retrained to detect other gestures

This demo runs on the Himax HM01BO IPduino Shield board.

Lattice sensAI

Jump to

Block Diagram

Face Detect Block Diagram


Quick Reference
Select All
Himax HM01B0 UPduino Shield Based Hand Gesture Detection Demonstration User Guide
1.0 9/26/2018 PDF 582.6 KB

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