Searching for MobileNetV3
By Andrew G. Howard, Mark Sandler, Grace Chu, Liang-Chieh Chen, Bo Chen, Mingxing Tan, Weijun Wang, Yukun Zhu, Ruoming Pang, Vijay Vasudevan and 2 more. 2019
9,744 citations recorded by the source corpus.
We present the next generation of MobileNets based on a combination of complementary search techniques as well as a novel architecture design. MobileNetV3 is tuned to mobile phone CPUs through a combination of hardware-aware network architecture search (NAS) complemented by the NetAdapt algorithm and then subsequently improved through novel architecture advances. This paper starts the exploration of how automated search algorithms and network design can work together to harness complementary approaches improving the overall state of the art. Through this process we create two new MobileNet models for release: MobileNetV3-Large and MobileNetV3-Small which are targeted for high and low resource use cases. These models are then adapted and applied to the tasks of object detection and semantic segmentation. For the task of semantic segmentation (or any dense pixel prediction), we propose a new efficient segmentation decoder Lite Reduced Atrous Spatial Pyramid Pooling (LR-ASPP). We achieve new state of the art results for mobile classification, detection and segmentation. MobileNetV3-Large is 3.2% more accurate on ImageNet classification while reducing latency by 20% compared to MobileNetV2. MobileNetV3-Small is 6.6% more accurate compared to a MobileNetV2 model with comparable latency. MobileNetV3-Large detection is over 25% faster at roughly the same accuracy as MobileNetV2 on COCO detection. MobileNetV3-Large LR-ASPP is 34% faster than MobileNetV2 R-ASPP at similar accuracy for Cityscapes segmentation.
Subfields: Computer Vision and Pattern Recognition, Electrical and Electronic Engineering, Efficiency & Compression
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Ming-Hsuan Yang
- Xiaogang Wang
- Trevor Darrell
- Alan L. Yuille
- Jitendra Malik
- Yi Yang
In Electrical and Electronic Engineering
- Carole-Jean Wu
- Young Geun Kim
- Song Han
- Lu Yuan
- Davide Scaramuzza
- Quoc V. Le
- Kurt Keutzer
- Tien-Ju Yang
In Efficiency & Compression
Citations in the corpus
- Chasing Carbon: The Elusive Environmental Footprint of Computing (2020, methodology): “For instance, when running on a CPU, MobileNet v2 is 17 × faster than Inception v3 [70].”
- Visual Transformers: Token-based Image Representation and Processing for Computer Vision (2020, neutral): “Early work in this direction includes [23, 31, 13, 17, 32, 16, 48, 27, 43].”
- Axial-DeepLab: Stand-Alone Axial-Attention for Panoptic Segmentation (2020, neutral): “Additionally, we multiply all the channels [36,74,35] by 0.”
- Reinforce Data, Multiply Impact: Improved Model Accuracy and Robustness with Dataset Reinforcement (2023, methodology): “To understand what makes a good teacher to reinforce datasets, we perform knowledge distillation with a variety of pretrained models in the Timm library [63] distilled to three representative student architectures MobileNetV3-large [25], ResNet-50 [21], and ViT-Small [16].”
- AutoGAN-Distiller: Searching to Compress Generative Adversarial Networks (2020, positive): “Already by now, NAS methods have outperformed manually designed architectures on a range of tasks such as image classification (Tan & Le, 2019; Tan et al., 2019; Howard et al., 2019; Liu et al., 2018a), and segmentation (Chen et al.”
- FBNetV3: Joint Architecture-Recipe Search using Neural Acquisition Function (2020, methodology): “This includes convolutional variants like depthwise convolutions in MobileNet, inverted residual blocks in MobileNetV2; activations such as hswish in MobileNetV3 [14, 31, 13]; and memory-based operations like shift [39] and shuffle [24].”
- BigNAS: Scaling Up Neural Architecture Search with Big Single-Stage Models (2020, methodology): “Following previous resource-aware NAS methods [6, 16, 32–34], our network architectures consist of a stack with inverted bottleneck residual blocks (MBConv) [29].”
- AtomNAS: Fine-Grained End-to-End Neural Architecture Search (2019, methodology): “Howard et al. (2019); Cai et al. (2019); Wu et al. (2019); Stamoulis et al. (2019a) search the expansion ratios in the MobileNetV2 block but still limit them to a few discrete values.”
- EfficientDet: Scalable and Efficient Object Detection (2019, background): “…In order to obtain better accuracy, it is common to scale up a baseline detector by employing bigger backbone networks ( e.g ., from mobile-size models [30, 10] and ResNet [9], to ResNeXt [33] and AmoebaNet [24]), or increasing input image size ( e.g ., from 512x512 [17] to 1536x1536 [37]).”
- Once for All: Train One Network and Specialize it for Efficient Deployment (2019, critical): “We use the same architecture space as MobileNetV3 (Howard et al., 2019).”
- FairNAS: Rethinking Evaluation Fairness of Weight Sharing Neural Architecture Search (2019, methodology): “7 MobileNetV3 Large [14] 219 5.”
- LambdaNetworks: Modeling Long-Range Interactions Without Attention (2021, methodology): “We use the same hyperparameters as Howard et al. (2019).”
- Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation (2019, methodology): “Note the channels in last block of MNV3 are reduced by a factor of 2 [27].”
- ResNeSt: Split-Attention Networks (2020, background): “Despite their superior computation and accuracy tradeoff in classification tasks, these models do not transfer well to other tasks as their isolated representations cannot capture cross-channel relationships [27,28].”
- Training with Quantization Noise for Extreme Model Compression (2020, methodology): “…on efficient parameterization, via weight pruning (LeCun et al., 1990; Li et al., 2016; Huang et al., 2018; Mittal et al., 2018), weight sharing (Dehghani et al., 2018; Turc et al., 2019; Lan et al., 2019) or with dedicated architectures (Tan & Le, 2019; Zhang et al., 2017; Howard et al., 2019).”
- Meta-Learning in Neural Networks: A Survey (2020, background): “Swish has gone on to contribute to several influential state-of-the-art and general purpose CNN architectures [235], [236].”
- ImageNet-21K Pretraining for the Masses (2021, background): “However, we show consistently and significantly that even small mobile-oriented models, like MobileNetV3 and OFA-595, can benefit from pretraining on a large (publicly available) dataset like ImageNet-21K-P.”
- MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices (2020, background): “…(CNN) for mobile vision applications (Iandola et al., 2016; Howard et al., 2017; Zhang et al., 2017, 2018; Sandler et al., 2018; Tan et al., 2019; Howard et al., 2019), they are usually tailored for CNN. Popular lightweight operations such as depth-wise convolution (Howard et al., 2017) cannot…”
- Adversarial Examples Improve Image Recognition (2019, neutral): “2% accuracy [11].”