Accelerating Binarized Neural Network Inference by Reusing Operation Results and Elevating Resource Utilization on Edge devices

Abstract

Convolutional Neural Networks (CNNs) have shown their abilities in computer vision such as self-driving car and image classification applications. However, computation, power consumption, and memory requirement remain challenges for CNNs when applied in the domains of edge devices. To address these challenges, one of the possible solutions is using Binarized Neural Networks (BNNs). Researchers have demonstrated that BNN models dramatically reduce computational complexity and memory requirements with acceptable accuracy loss. This paper presents an accelerator design for BNN inference that minimizes the number of operations and elevates resource utilization on edge devices. We implemented our accelerator on the Xilinx ZCU104 FPGA and evaluated it with VGG-13 like BNN networks for Tiny ImageNet. Experimental results show that our accelerator can reduce, on average, 99.6% operations and achieve up to 1974 speedup on an FPGA platform compared to state-of-the-art design.