zmswanson / fpga-vision

fpga-vision

Goal: Perform optical character recognition (OCR) on simple character set (0-9) entirely on FPGA.

This project was inspired by the computer vision demo on the Hello FPGA Kit from Microchip (Microsemi).

Initial setup:

• Digilent Basys 3 board with a Xilinx Artix-7 FPGA
• Omnivision OV7670 640x480 CMOS camera
• TFT SPI OLED display or VGA display

Goals:

• Interface FPGA with OV7670 camera
• Control VGA display
• Display camera data on VGA display
• Control SPI OLED display
• Display camera data on SPI OLED display
• Draw overlays on (e.g. bounding boxes) on camera data
• Develop CNN logic for FPGA
• Train OCR model in Python
• Port Python-trained weights to FPGA CNN
• Translate CNN output to numeric overlays for image
• Port logic designs to Intel Cyclone 10 architecture