MobileNetV2: Inverted Residuals and Linear Bottlenecks
In this paper we describe a new mobile architecture, MobileNetV2, that improves the state of the art performance of mobile models on multiple tasks and benchmarks as well as across a spectrum of different model sizes. We also describe efficient ways of applying these mobile models to object detection in a novel framework we call SSDLite.
Also cited · not yet reviewed (5)
- MobileNets2017 · cited 8×, 6 in Method“Our network design is based on MobileNetV1 MobilenetV1.”From this paper · §Related Work
- ResNet2015 · cited 5×, 2 in Method“Both manual architecture search and improvements in training algorithms, carried out by numerous teams has lead to dramatic improvements over early designs such as AlexNet AlexNet, VGGNet VGGNet, GoogLeNet GoogleNet. , a…”From this paper · §Related Work
- ResNeXt2016 · cited 2×, 1 in Method“When the expansion ratio is smaller than 11, this is a classical residual convolutional block ResNet; ResNext2016.”From this paper · §Preliminaries, discussion and intuition
- VGG2014 · cited 2דBoth manual architecture search and improvements in training algorithms, carried out by numerous teams has lead to dramatic improvements over early designs such as AlexNet AlexNet, VGGNet VGGNet, GoogLeNet GoogleNet. , a…”From this paper · §Related Work
Show 1 more
- NASNet2017 · cited 2דRecently, LearningToLearnScale; GeneticCNN; EvolutionImageClassifiers; NAS_reinforcement, opened up a new direction of bringing optimization methods including genetic algorithms and reinforcement learning to architectura…”From this paper · §Related Work
Led to
- MnasNet2018 · cited 10×, 3 in Method“Compared to the MobileNetV2 mobilenetv218, our model improves the ImageNet accuracy by 3.0% with similar latency on the Google Pixel phone.”From MnasNet · §Introduction
- EfficientNet2019 · cited 5×, 2 in Method“Scaling network width is commonly used for small size models (Howard et al. 2017; Sandler et al. 2018; Tan et al. 2019)22 2 In some literature, scaling number of channels is called “depth multiplier”, which means the sam…”From EfficientNet · §Compound Model Scaling
- CoAtNet2021 · cited 3×, 1 in Method“Traditionally, regular convolutions, such as ResNet blocks [3], are popular in large-scale ConvNets; in contrast, depthwise convolutions [28] are popular in mobile platforms due to its lower computational cost and smalle…”From CoAtNet · §Related Work
Abstract
In this paper we describe a new mobile architecture, MobileNetV2, that improves the state of the art performance of mobile models on multiple tasks and benchmarks as well as across a spectrum of different model sizes. We also describe efficient ways of applying these mobile models to object detection in a novel framework we call SSDLite. Additionally, we demonstrate how to build mobile semantic segmentation models through a reduced form of DeepLabv3 which we call Mobile DeepLabv3. The MobileNetV2 architecture is based on an inverted residual structure where the input and output of the residual block are thin bottleneck layers opposite to traditional residual models which use expanded representations in the input an MobileNetV2 uses lightweight depthwise convolutions to filter features in the intermediate expansion layer. Additionally, we find that it is important to remove non-linearities in the narrow layers in order to maintain representational power. We demonstrate that this improves performance and provide an intuition that led to this design. Finally, our approach allows decoupling of the input/output domains from the expressiveness of the transformation, which provides a convenient framework for further analysis. We measure our performance on Imagenet classification, COCO object detection, VOC image segmentation. We evaluate the trade-offs between accuracy, and number of operations measured by multiply-adds (MAdd), as well as the number of parameters