MnasNet: Platform-Aware Neural Architecture Search for Mobile

CVPR 2019 Mingxing TanBo ChenRuoming PangVijay VasudevanMark SandlerAndrew HowardQuoc V. Le

Designing convolutional neural networks (CNN) for mobile devices is challenging because mobile models need to be small and fast, yet still accurate. Although significant efforts have been dedicated to design and improve mobile CNNs on all dimensions, it is very difficult to manually balance these trade-offs when there are so many architectural possibilities to consider... (read more)

PDF Abstract

Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT LEADERBOARD
Real-Time Object Detection COCO MobileNetV2 + SSDLite MAP 22.1 # 13
Image Classification ImageNet MnasNet-A3 Top 1 Accuracy 76.7% # 105
Top 5 Accuracy 93.3% # 73
Number of params 5.2M # 2
Image Classification ImageNet MnasNet-A2 Top 1 Accuracy 75.6% # 117
Top 5 Accuracy 92.7% # 82
Number of params 4.8M # 2
Image Classification ImageNet MnasNet-A1 Top 1 Accuracy 75.2% # 121
Top 5 Accuracy 92.5% # 86
Number of params 3.9M # 2