nn benchmark

nn benchmark数据集是由布列塔尼国立高等先进技术学院的研究人员创建，用于评估量化神经网络在FPGA上的性能。该数据集包含49个不同量化配置的神经网络实现，主要用于图像识别任务，特别是FashionMNIST数据集。创建过程中，研究人员使用了Xilinx的FINN和Brevitas框架来训练和部署量化神经网络。该数据集的应用领域主要集中在优化神经网络在低功耗硬件上的部署，旨在解决资源受限环境下的高效计算问题。

The nn benchmark dataset was created by researchers from École Nationale Supérieure de Techniques Avancées de Bretagne (ENSTA Bretagne) to evaluate the performance of quantized neural networks on field-programmable gate arrays (FPGAs). This dataset includes 49 neural network implementations with distinct quantization configurations, primarily designed for image recognition tasks, especially those based on the FashionMNIST dataset. During its development, researchers utilized Xilinx's FINN and Brevitas frameworks for training and deploying the quantized neural networks. The main application scope of this dataset focuses on optimizing the deployment of neural networks on low-power hardware, aiming to address the challenge of efficient computing in resource-constrained environments.