📝 This post was initially released on the HLS Works Blog in 2020. The post was moved to this website after HLS Works closed in Sep 2021. Project showcase - CNN layer library in HLS The main goal of this project was to create an HLS library for implementing CNN’s convolutional layers. Our client used this library to study FPGA performance against other alternatives. Pixel per clock per clock: The generated IP core is able to process one pixel per clock 99% of the time.