- LEVEL ACCURACY WITH 50 X FEWER PARAMETERS AND < 0 . 5 MB MODEL SIZE
- LEVEL ACCURACY WITH 50 X FEWER PARAMETERS AND < 0 . 5 MB MODEL SIZE
复制标题
DOI：
--
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
F. Iandola;Song Han;Matthew W. Moskewicz;Khalid Ashraf;W. Dally;K. Keutzer
中科院分区：
文献类型：
--
作者：
F. Iandola;Song Han;Matthew W. Moskewicz;Khalid Ashraf;W. Dally;K. Keutzer
最近对深度卷积神经网络（CNN）的研究主要集中在提高准确性上。对于给定的准确度水平，通常可以识别实现该准确度水平的多个CNN架构。在同等精度的情况下，较小的CNN架构至少提供了三个优势：（1）较小的CNN在分布式训练期间需要较少的服务器通信。(2)较小的CNN需要更少的带宽将新模型从云端导出到自动驾驶汽车。(3)较小的CNN更适合部署在FPGA和其他内存有限的硬件上。为了提供所有这些优点，我们提出了一种称为SqueezeNet的小型CNN架构。SqueezeNet在ImageNet上实现了AlexNet级别的准确性，参数减少了50倍。此外，通过模型压缩技术，我们能够将SqueezeNet压缩到0.5MB以下（比AlexNet小510倍）。SqueezeNet架构可在此处下载：https://github.com/DeepScale/SqueezeNet
Recent research on deep convolutional neural networks (CNNs) has focused primarily on improving accuracy. For a given accuracy level, it is typically possible to identify multiple CNN architectures that achieve that accuracy level. With equivalent accuracy, smaller CNN architectures offer at least three advantages: (1) Smaller CNNs require less communication across servers during distributed training. (2) Smaller CNNs require less bandwidth to export a new model from the cloud to an autonomous car. (3) Smaller CNNs are more feasible to deploy on FPGAs and other hardware with limited memory. To provide all of these advantages, we propose a small CNN architecture called SqueezeNet. SqueezeNet achieves AlexNet-level accuracy on ImageNet with 50x fewer parameters. Additionally, with model compression techniques, we are able to compress SqueezeNet to less than 0.5MB (510× smaller than AlexNet). The SqueezeNet architecture is available for download here: https://github.com/DeepScale/SqueezeNet