GhostNet: More Features from Cheap Operations

Abstract

Deploying convolutional neural networks (CNNs) on embedded devices is difficult due to the limited memory and computation resources. The redundancy in feature maps is an important characteristic of those successful CNNs, but has rarely been investigated in neural architecture design. This paper proposes a novel Ghost module to generate more feature maps from cheap operations. Based on a set of intrinsic feature maps, we apply a series of linear transformations with cheap cost to generate many ghost feature maps that could fully reveal information underlying intrinsic features. The proposed Ghost module can be taken as a plug-and-play component to upgrade existing convolutional neural networks. Ghost bottlenecks are designed to stack Ghost modules, and then the lightweight GhostNet can be easily established. Experiments conducted on benchmarks demonstrate that the proposed Ghost module is an impressive alternative of convolution layers in baseline models, and our GhostNet can achieve higher recognition performance (e.g. $75.7\%$ top-1 accuracy) than MobileNetV3 with similar computational cost on the ImageNet ILSVRC-2012 classification dataset. Code is available at https://github.com/huawei-noah/ghostnet

arXiv commentCVPR 2020. Code is available at https://github.com/huawei-noah/ghostnet

References in corpus (13)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• Distilling the Knowledge in a Neural Network
• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
• SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size
• Exploiting Linear Structure Within Convolutional Networks for Efficient Evaluation
• Rethinking the Value of Network Pruning
• FasterSeg: Searching for Faster Real-time Semantic Segmentation
• MetaPruning: Meta Learning for Automatic Neural Network Channel Pruning
• Model Compression with Adversarial Robustness: A Unified Optimization Framework
• CARS: Continuous Evolution for Efficient Neural Architecture Search
• Positive-Unlabeled Compression on the Cloud
• All You Need is a Few Shifts: Designing Efficient Convolutional Neural Networks for Image Classification

Cited by in corpus (14)

• Bringing AI To Edge: From Deep Learning's Perspective
• Compacting Deep Neural Networks for Internet of Things: Methods and Applications
• CARS: Continuous Evolution for Efficient Neural Architecture Search
• AdaFuse: Adaptive Temporal Fusion Network for Efficient Action Recognition
• MixPath: A Unified Approach for One-shot Neural Architecture Search
• Rethinking Bottleneck Structure for Efficient Mobile Network Design
• A Semi-Supervised Assessor of Neural Architectures
• Structured Convolutions for Efficient Neural Network Design
• A Novel lightweight Convolutional Neural Network, ExquisiteNetV2
• DC-NAS: Divide-and-Conquer Neural Architecture Search
• Universal Adder Neural Networks
• EADNet: Efficient Asymmetric Dilated Network for Semantic Segmentation
• A Feature-map Discriminant Perspective for Pruning Deep Neural Networks
• Uniqueness and uniform structural stability of Poiseuille flows in an infinitely long pipe with Navier boundary conditions