This reference design provides an example of how to implement a face/object tracking application based on the Lattice sensAI stack. It includes a SPI, DDR memory interface IP blocks, an image signal processing engine, 8 CNN acceleration engines and an overlay engine to indicate the face in the image.
The reference design is optimized for the ECP5-85K FPGA. While the FPGA logic will be loaded from the external flash memory, the weights and biases are loaded from a SD card, connected to SPI interface.
Features
- Green box indicates location of face in image
- Runs @ 14fps with 90 x 90 RGB input
- Consumes 0.85W of power consumption on ECP5 FPGA
- Applications include selfie drones, surveillance cameras, smart cameras etc
