Implementation of CNNs on FPGA

Charge | 2020

CNNs require a good computing power. So, GPUs are used to meet the requirements. GPUs have some disadvantages – High power consumption and lack of ports(I/Os). Many applications like drones, Embedded systems, self-driving cars etc which can be highly benefited from DL algos. But they need to work on low power. FPGAs can be used for this purpose to overcome the limitations of GPUs. High end FPGAs are able to provide comparable or even better performance than state of art GPUs. Many CNN models are implemented using well-known python frameworks like keras, tensorflow, caffe, pytorch etc.. To implement these on FPGA one should deal with a lot of hardware design. This project is aimed at creating a tool which can synthesize a DL inference model which can be deployed on FPGA.

deep learning