用FPGA加速图像分类模型，性能远超CPU/GPU。
FPGA-based Acceleration of Neural Network for Image Classification using Vitis AI
- 在ZCU104 FPGA上用Vitis AI加速CNN模型
- 吞吐量提升3.33-5.82倍，能效提升3.39-6.30倍
- 适合边缘设备部署，可支持深度估计等下游任务
近年来，卷积神经网络（CNN）在计算机视觉中广泛应用。在CPU或GPU上运行复杂CNN架构时，常面临吞吐量不足或功耗过高的问题。为此，亟需专用硬件加速计算负载以突破这些限制。本文在Xilinx Zynq UltraScale+ MPSoC ZCU104 FPGA评估板上，使用Vitis AI加速针对CIFAR-10数据集的CNN图像分类模型。实验结果表明，相比CPU和GPU基线，该方案实现3.33-5.82倍的吞吐量提升，以及3.39-6.30倍的能效提升。研究展示了该方法在提取2D特征用于后续任务（如深度估计与3D重建）方面的潜力。
原文摘要 · Abstract (English)
In recent years, Convolutional Neural Networks (CNNs) have been widely adopted in computer vision. Complex CNN architecture running on CPU or GPU has either insufficient throughput or prohibitive power consumption. Hence, there is a need to have dedicated hardware to accelerate the computation workload to solve these limitations. In this paper, we accelerate a CNN for image classification with the CIFAR-10 dataset using Vitis-AI on Xilinx Zynq UltraScale+ MPSoC ZCU104 FPGA evaluation board. The work achieves 3.33-5.82x higher throughput and 3.39-6.30x higher energy efficiency than CPU and GPU baselines. It shows the potential to extract 2D features for downstream tasks, such as depth estimation and 3D reconstruction.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。