Blog

The latest from Google Research

EfficientNet: Improving Accuracy and Efficiency through AutoML and Model Scaling

Wednesday, May 29, 2019

Posted by Mingxing Tan, Staff Software Engineer and Quoc V. Le, Principal Scientist, Google AIConvolutional neural networksResNetGPipeImageNetICML 2019EfficientNet: Rethinking Model Scaling for Convolutional Neural Networkscompound coefficientAutoMLCompound Model Scaling: A Better Way to Scale Up CNNsgrid searchFLOPS

Comparison of different scaling methods. Unlike conventional scaling methods (b)-(d) that arbitrary scale a single dimension of the network, our compound scaling method uniformly scales up all dimensions in a principled way.

MobileNetResNetEfficientNet Architectureneural architecture searchAutoML MNAS frameworkMobileNetV2MnasNetEfficientNets

The architecture for our baseline network EfficientNet-B0 is simple and clean, making it easier to scale and generalize.

EfficientNet Performance ImageNetGpipeResNet-50

Model Size vs. Accuracy Comparison. EfficientNet-B0 is the baseline network developed by AutoML MNAS, while Efficient-B1 to B7 are obtained by scaling up the baseline network. In particular, our EfficientNet-B7 achieves new state-of-the-art 84.4% top-1 / 97.1% top-5 accuracy, while being 8.4x smaller than the best existing CNN.

CIFAR-100FlowersTPUhereAcknowledgements:Special thanks to Hongkun Yu, Ruoming Pang, Vijay Vasudevan, Alok Aggarwal, Barret Zoph, Xianzhi Du, Xiaodan Song, Samy Bengio, Jeff Dean, and the Google Brain team.