Face Detection – Software to Hardware using High Level Synthesis

Abstract

In this article, we demonstrated a concept for converting OpenCV based software implementation of face detection application [1] to hardware implementation using High Level Synthesis tool. With the help of High Level Synthesis tool and certain structural optimization at system level we are able to generate optimized parallel and pipelined hardware of whole face detection application which can give exactly same accuracy as compare to standard software implementation with better speed. HLS tool helps us in generating efficient gate level design of complete face detection application with gate area of 300k at 200 MHz and processes 8 images (size 800x600) per second. We managed to process image stream in pipelined manner at each stage like color conversion, image scaling and histogram equalization. We also managed to calculate different feature on sub window size of 20x20 in pipelined manner for better efficiency.