EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks

ICML 2019 Mingxing TanQuoc V. Le

Convolutional Neural Networks (ConvNets) are commonly developed at a fixed resource budget, and then scaled up for better accuracy if more resources are available. In this paper, we systematically study model scaling and identify that carefully balancing network depth, width, and resolution can lead to better performance... (read more)

PDF Abstract

Code


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK USES EXTRA
TRAINING DATA
RESULT LEADERBOARD
Fine-Grained Image Classification Birdsnap EfficientNet-B7 Accuracy 84.3% # 1
Image Classification CIFAR-10 EfficientNet-B7 Percentage correct 98.9 # 2
PARAMS 64M # 7
Image Classification CIFAR-100 EfficientNet-B7 Percentage correct 91.7 # 3
PARAMS 64M # 7
Fine-Grained Image Classification FGVC Aircraft EfficientNet-B7 Accuracy 92.9% # 11
Image Classification Flowers-102 EfficientNet-B7 Accuracy 98.8% # 4
Fine-Grained Image Classification Food-101 EfficientNet-B7 Accuracy 93.0 # 1
Image Classification ImageNet EfficientNet-B7 Top 1 Accuracy 84.4% # 21
Top 5 Accuracy 97.1% # 15
Number of params 66M # 21
Image Classification ImageNet EfficientNet-B5 Top 1 Accuracy 83.3% # 30
Top 5 Accuracy 96.7% # 19
Number of params 30M # 33
Image Classification ImageNet EfficientNet-B3 Top 1 Accuracy 81.1% # 51
Top 5 Accuracy 95.5% # 33
Number of params 12M # 44
Image Classification ImageNet EfficientNet-B2 Top 1 Accuracy 79.8% # 66
Top 5 Accuracy 94.9% # 41
Number of params 9.2M # 46
Image Classification ImageNet EfficientNet-B1 Top 1 Accuracy 78.8% # 79
Top 5 Accuracy 94.4% # 56
Number of params 7.8M # 49
Image Classification ImageNet EfficientNet-B0 Top 1 Accuracy 76.3% # 111
Top 5 Accuracy 93.2% # 76
Number of params 5.3M # 59
Image Classification ImageNet EfficientNet-B6 Top 1 Accuracy 84% # 25
Top 5 Accuracy 96.9% # 17
Number of params 43M # 26
Image Classification ImageNet EfficientNet-B4 Top 1 Accuracy 82.6% # 37
Top 5 Accuracy 96.3% # 23
Number of params 19M # 43
Fine-Grained Image Classification Oxford-IIIT Pets EfficientNet-B7 Accuracy 95.4% # 2
Fine-Grained Image Classification Stanford Cars EfficientNet-B7 Accuracy 94.7% # 9

Methods used in the Paper