MobileOne: An Improved One millisecond Mobile Backbone

Efficient neural network backbones for mobile devices are often optimized for metrics such as FLOPs or parameter count. However, these metrics may not correlate well with latency of the network when deployed on a mobile device. Therefore, we perform extensive analysis of different metrics by deploying several mobile-friendly networks on a mobile device. We identify and analyze architectural and optimization bottlenecks in recent efficient neural networks and provide ways to mitigate these bottlenecks. To this end, we design an efficient backbone MobileOne, with variants achieving an inference time under 1 ms on an iPhone12 with 75.9% top-1 accuracy on ImageNet. We show that MobileOne achieves state-of-the-art performance within the efficient architectures while being many times faster on mobile. Our best model obtains similar performance on ImageNet as MobileFormer while being 38x faster. Our model obtains 2.3% better top-1 accuracy on ImageNet than EfficientNet at similar latency. Furthermore, we show that our model generalizes to multiple tasks - image classification, object detection, and semantic segmentation with significant improvements in latency and accuracy as compared to existing efficient architectures when deployed on a mobile device. Code and models are available at https://github.com/apple/ml-mobileone

PDF Abstract CVPR 2023 PDF CVPR 2023 Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Classification ImageNet MobileOne-S4 (distill) Top 1 Accuracy 81.4% # 590
Number of params 14.8M # 512
GFLOPs 2.9 # 171
Image Classification ImageNet MobileOne-S3 (distill) Top 1 Accuracy 80.0% # 668
Number of params 10.1M # 475
GFLOPs 1.8 # 138
Image Classification ImageNet MobileOne-S2 (distill) Top 1 Accuracy 79.1% # 718
Number of params 7.8M # 459
GFLOPs 1.3 # 118
Image Classification ImageNet MobileOne-S1 (distill) Top 1 Accuracy 77.4% # 813
Number of params 4.8M # 394
GFLOPs 0.825 # 96
Image Classification ImageNet MobileOne-S0 (distill) Top 1 Accuracy 72.5% # 927
Number of params 2.1M # 357
GFLOPs 0.275 # 22
Image Classification ImageNet MobileOne-S0 Top 1 Accuracy 71.4% # 941
Number of params 2.1M # 357
GFLOPs 0.275 # 22
Image Classification ImageNet MobileOne-S1 Top 1 Accuracy 75.9 # 861
Number of params 4.8M # 394
GFLOPs 0.825 # 96
Image Classification ImageNet MobileOne-S2 Top 1 Accuracy 77.4% # 813
Number of params 7.8 # 1
GFLOPs 1.3 # 118
Image Classification ImageNet MobileOne-S3 Top 1 Accuracy 78.1% # 789
Number of params 10.1M # 475
GFLOPs 1.8 # 138
Image Classification ImageNet MobileOne-S4 Top 1 Accuracy 79.4% # 699
Number of params 14.8M # 512
GFLOPs 2.9 # 171

Methods