GhostNet: More Features From Cheap Operations

Han, Kai; Wang, Yunhe; Tian, Qi; Guo, Jianyuan; Xu, Chunjing; Xu, Chang

doi:10.1109/cvpr42600.2020.00165

Cited by 2,036 publications

(1,000 citation statements)

References 28 publications

Supporting

Mentioning

957

Contrasting

Order By: Relevance

“…The simple linear cheap operation is also used to generate more feature information. In the case of the same human visual perception, the size of the ghost network convolution kernel is 1 × 1 in our work compared with other convolution kernels of 3 × 3 and 5 × 5 [ 46 ]. This is conducive to the extraction of the local features of the images, but fewer parameters are calculated.…”

Section: Experimental Results and Analysismentioning

confidence: 99%

“…The ghost network is introduced into DGRAMN in this article. The total number of parameters and computational complexity required in the ghost network is reduced compared with an ordinary CNN without changing the size of the output feature map [ 46 ]. The use of GhostNet solves the problem of large computation and large storage in spectral reconstruction.…”

Section: Our Methodsmentioning

confidence: 99%

“…We set the convolution kernel size of this layer to 1 × 1 to simplify the calculation and do not use bias in the ordinary convolution operation (Equation (3)). A linear operation [ 46 ] is used to generate multiple feature images. The generated feature graph is then normalized, which not only retains the original learning features but also accelerates the running time of the hardware training data.…”

Section: Our Methodsmentioning

confidence: 99%

See 2 more Smart Citations

Double Ghost Convolution Attention Mechanism Network: A Framework for Hyperspectral Reconstruction of a Single RGB Image

Wang

Jiangwei

2021

Sensors

View full text Add to dashboard Cite

Current research on the reconstruction of hyperspectral images from RGB images using deep learning mainly focuses on learning complex mappings through deeper and wider convolutional neural networks (CNNs). However, the reconstruction accuracy of the hyperspectral image is not high and among other issues the model for generating these images takes up too much storage space. In this study, we propose the double ghost convolution attention mechanism network (DGCAMN) framework for the reconstruction of a single RGB image to improve the accuracy of spectral reconstruction and reduce the storage occupied by the model. The proposed DGCAMN consists of a double ghost residual attention block (DGRAB) module and optimal nonlocal block (ONB). DGRAB module uses GhostNet and PRELU activation functions to reduce the calculation parameters of the data and reduce the storage size of the generative model. At the same time, the proposed double output feature Convolutional Block Attention Module (DOFCBAM) is used to capture the texture details on the feature map to maximize the content of the reconstructed hyperspectral image. In the proposed ONB, the Argmax activation function is used to obtain the region with the most abundant feature information and maximize the most useful feature parameters. This helps to improve the accuracy of spectral reconstruction. These contributions enable the DGCAMN framework to achieve the highest spectral accuracy with minimal storage consumption. The proposed method has been applied to the NTIRE 2020 dataset. Experimental results show that the proposed DGCAMN method outperforms the spectral accuracy reconstructed by advanced deep learning methods and greatly reduces storage consumption.

show abstract

Section: Experimental Results and Analysismentioning

confidence: 99%

Section: Our Methodsmentioning

confidence: 99%

Section: Our Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Double Ghost Convolution Attention Mechanism Network: A Framework for Hyperspectral Reconstruction of a Single RGB Image

Wang

Jiangwei

2021

Sensors

View full text Add to dashboard Cite

show abstract

“…The GhostNet [27] idea comes from Huawei's Noah's Ark Laboratory. It proposes that a well-trained deep neural network usually contains rich or even redundant feature maps, and one feature map can be transformed from another feature map through certain operations.…”

Section: Related Workmentioning

confidence: 99%

“…In the refined design of compression, the deep separable convolution in MobileNet was first used to reduce the parameters of the network initially, however it did not reach the ideal goal of this article. GhostNet [27] is also a kind of compact model design. It proposes that a feature map can be used as a "Ghost" of another and the "Ghost" can be generated by a cheaper operation.…”

Section: Introductionmentioning

confidence: 99%

Real-Time Segmentation Method of Lightweight Network For Finger Vein Using Embedded Terminal Technique

et al. 2021

View full text Add to dashboard Cite

A Lightweight YOLO Object Detection Algorithm Based on Bidirectional Multi‐Scale Feature Enhancement

Liu,

Zhang,

Zhang

et al. 2024

Advcd Theory and Sims

View full text Add to dashboard Cite

This paper proposes a lightweight YOLO object detection algorithm based on bidirectional multi‐scale feature enhancement. The problem is that the original YOLOv5 algorithm does not make full use of the relationship between the feature layers, resulting in the loss of target semantic information and a large number of parameters. First, a bidirectional multi‐scale feature‐enhanced weighted fusion backbone network is constructed to extract target features repeatedly. It enhances the fusion ability of shallow detail features and high‐level semantic information to capture richer multi‐scale semantic information. Second, the NCA attention module is built and integrated into the feature fusion network to enhance the critical characteristics of the target region. Finally, the Ghost module is used instead of the convolutional blocks in the original network to lighten the model while reducing the network complexity and training difficulty. Experimental results show that the improved YOLOv5 algorithm achieves 78.8% mAP@0.5 for the PASCAL VOC2012 dataset, which is 1.5% higher than the original algorithm, at 62.5 FPS. The number of parameters is also reduced by 43.6%. The mAP@0.5 on the self‐made metal foreign object dataset reached 98.4%, at 58.8 FPS, which can meet the requirements of end‐device deployment and real‐time detection.

show abstract

GhostNet: More Features From Cheap Operations

Cited by 2,036 publications

References 28 publications

Double Ghost Convolution Attention Mechanism Network: A Framework for Hyperspectral Reconstruction of a Single RGB Image

Double Ghost Convolution Attention Mechanism Network: A Framework for Hyperspectral Reconstruction of a Single RGB Image

Real-Time Segmentation Method of Lightweight Network For Finger Vein Using Embedded Terminal Technique

A Lightweight YOLO Object Detection Algorithm Based on Bidirectional Multi‐Scale Feature Enhancement

Contact Info

Product

Resources

About