Customizable to detect hand gestures: This design implements a low power AI based system to detect hand gestures using an IR image sensor.
Optimized Power and Performance – iCE40 UltraPlus low power FPGA running the CNN compact accelerator engine is used for processing the images consuming few milli-watts of power.
Complete Reference Design - Training dataset, scripts for training using common Neural Network training tools and NN model are provided to enable modification.
Features
- AI based on-device gesture detection processing
- VGG like 8-bits CNN used inside of iCE40 UltraPlus at up to 10 frames per second
- Adjustable Frame-rate. Can be optimized between power and response time depending on system needs
- Gestures can be customized based on system needs
