SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size

Forrest N. Iandola, Song Han, Matthew W. Moskewicz, Khalid Ashraf, William J. Dally +1 more
2/24/2016
cs.CVcs.AI

Official code: forresti/SqueezeNet (2,221 stars, BSD-2-Clause, last commit 2018-07-09). 4 more implementations below.

Abstract

Recent research on deep neural networks has focused primarily on improving accuracy. For a given accuracy level, it is typically possible to identify multiple DNN architectures that achieve that accuracy level. With equivalent accuracy, smaller DNN architectures offer at least three advantages: (1) Smaller DNNs require less communication across servers during distributed training. (2) Smaller DNNs require less bandwidth to export a new model from the cloud to an autonomous car. (3) Smaller DNNs are more feasible to deploy on FPGAs and other hardware with limited memory. To provide all of these advantages, we propose a small DNN 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 (510x smaller than AlexNet). The SqueezeNet architecture is available for download here: https://github.com/DeepScale/SqueezeNet

View on arXivView PDF

Code Implementations(5)

forresti/SqueezeNetOfficial100%
BSD-2-Clause

SqueezeNet: AlexNet-level accuracy with 50x fewer parameters

2,221719Feb 22, 20168 years agoBSD-2-Clause
DT42/squeezenet_demoOfficial100%
GPL-3.0
23979PythonJun 3, 20168 years agoGPL-3.0
Element-Research/dpnnOfficial100%
BSD-3-Clause

deep extensions to nn

19284Lua, CMakeApr 6, 20159 years agoBSD-3-Clause
ejlb/squeezenet-chainerOfficial100%

Chainer squeezenet model

244PythonFeb 26, 201610 years ago
hariag/SqueezeNetOfficial100%

SqueezeNet: AlexNet-level accuracy with 50x fewer parameters

215Shell, PythonMar 2, 201610 years ago

Similar Papers

TResNet: High Performance GPU-Dedicated Architecture

86% match2 repos36.9k2020

EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks

85% match1 repo5.3k2019

ZeRO: Memory Optimizations Toward Training Trillion Parameter Models

85% match2 repos43.2k2019

Model Rubik's Cube: Twisting Resolution, Depth and Width for TinyNets

84% match1 repo4.4k2020

Ristretto: Hardware-Oriented Approximation of Convolutional Neural Networks

83% match1 repo2.2k2016

Revisiting Data Augmentation in Model Compression: An Empirical and Comprehensive Study

83% match2023

Cite this paper

@article{iandola2016squeezenet,
  title  = {SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size},
  author = {Forrest N. Iandola and Song Han and Matthew W. Moskewicz and Khalid Ashraf and William J. Dally and Kurt Keutzer},
  year   = {2016},
  eprint = {1602.07360},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1602.07360}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1602.07360.svg)](https://codeofpaper.com/paper/1602.07360)

Discussion