Blog

The latest news from Google AI

MobileNetV2: The Next Generation of On-Device Computer Vision Networks

Tuesday, April 3, 2018

Posted by Mark Sandler and Andrew Howard, Google ResearchMobileNetV1MobileNetV2TensorFlow-Slim Image Classification LibraryColaboratorydownloadJupytermoduleson github¹

Overview of MobileNetV2 Architecture. Blue blocks represent composite convolutional building blocks as shown above.

MobileNet V2: Inverted Residuals and Linear BottlenecksHow does it compare to the first generation of MobileNets?

MobileNetV2 improves speed (reduced latency) and increased ImageNet Top 1 accuracy

Tensorflow Object Detection API

Model	Params	Multiply-Adds	mAP	Mobile CPU
MobileNetV1 + SSDLite	5.1M	1.3B	22.2%	270ms
MobileNetV2 + SSDLite	4.3M	0.8B	22.1%	200ms

announced recentlyPASCAL VOC 2012

Model	Params	Multiply-Adds	mIOU
MobileNetV1 + DeepLabV3	11.15M	14.25B	75.29%
MobileNetV2 + DeepLabV3	2.11M	2.75B	75.32%

Acknowledgements:We would like to acknowledge our core contributors Menglong Zhu, Andrey Zhmoginov and Liang-Chieh Chen. We also give special thanks to Bo Chen, Dmitry Kalenichenko, Skirmantas Kligys, Mathew Tang, Weijun Wang, Benoit Jacob, George Papandreou, Zhichao Lu, Vivek Rathod, Jonathan Huang, Yukun Zhu, and Hartwig Adam.References

MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications, Howard AG, Zhu M, Chen B, Kalenichenko D, Wang W, Weyand T, Andreetto M, Adam H, arXiv:1704.04861, 2017.

MobileNetV2: Inverted Residuals and Linear Bottlenecks, Sandler M, Howard A, Zhu M, Zhmoginov A, Chen LC. arXiv preprint. arXiv:1801.04381, 2018.

Rethinking Atrous Convolution for Semantic Image Segmentation, Chen LC, Papandreou G, Schroff F, Adam H. arXiv:1706.05587, 2017.

Speed/accuracy trade-offs for modern convolutional object detectors, Huang J, Rathod V, Sun C, Zhu M, Korattikara A, Fathi A, Fischer I, Wojna Z, Song Y, Guadarrama S, Murphy K, CVPR 2017.

Deep Residual Learning for Image Recognition, He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. arXiv:1512.03385,2015

1 The shortcut (also known as skip) connections, popularized by ResNets[5] are commonly used to connect the non-bottleneck layers. MobilenNetV2 inverts this notion and connects the bottlenecks directly.^↩