Multiple Feature Reweight DenseNet for Image Classification

Zhang, Ke; Guo, Yurong; Wang, Xinsheng; Yuan, Jinsha; Ding, Qin

doi:10.1109/access.2018.2890127

Cited by 176 publications

(61 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We refer to this network architecture as Dense because all the layers are connected to one another (dense connectivity). The superior performance of DenseNets over standard CNNs has been previously reported in the field of image learning and classification [37][38][39] . Likewise, in the present study on the processing of DP images, the proposed MSDN is expected to create rich patterns while maintaining the low complexity of information, thus enabling better classification performance.…”

Section: Shaped Dps In a Multistream Densenetmentioning

confidence: 74%

Identification of crystal symmetry from noisy diffraction patterns by a shape analysis and deep learning

et al. 2020

View full text Add to dashboard Cite

The robust and automated determination of crystal symmetry is of utmost importance in material characterization and analysis. Recent studies have shown that deep learning (DL) methods can effectively reveal the correlations between X-ray or electron-beam diffraction patterns and crystal symmetry. Despite their promise, most of these studies have been limited to identifying relatively few classes into which a target material may be grouped. On the other hand, the DL-based identification of crystal symmetry suffers from a drastic drop in accuracy for problems involving classification into tens or hundreds of symmetry classes (e.g., up to 230 space groups), severely limiting its practical usage. Here, we demonstrate that a combined approach of shaping diffraction patterns and implementing them in a multistream DenseNet (MSDN) substantially improves the accuracy of classification. Even with an imbalanced dataset of 108,658 individual crystals sampled from 72 space groups, our model achieves 80.12 ± 0.09% space group classification accuracy, outperforming conventional benchmark models by 17–27 percentage points (%p). The enhancement can be largely attributed to the pattern shaping strategy, through which the subtle changes in patterns between symmetrically close crystal systems (e.g., monoclinic vs. orthorhombic or trigonal vs. hexagonal) are well differentiated. We additionally find that the MSDN architecture is advantageous for capturing patterns in a richer but less redundant manner relative to conventional convolutional neural networks. The proposed protocols in regard to both input descriptor processing and DL architecture enable accurate space group classification and thus improve the practical usage of the DL approach in crystal symmetry identification.

show abstract

Section: Shaped Dps In a Multistream Densenetmentioning

confidence: 74%

Identification of crystal symmetry from noisy diffraction patterns by a shape analysis and deep learning

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Thus, combined with the corresponding agricultural knowledge, images of healthy and diseased leaves can be used as the input of CNN to train the identification model. Methods have been applied to a variety of food and cash crops including but not limited to rice [12,13], corn [14], tea [15][16][17], cannabis [18], and apple [19].…”

Section: Related Workmentioning

confidence: 99%

A Sparse Deep Transfer Learning Model and Its Application for Smart Agriculture

Chen

Zhang

Chen

et al. 2021

Wireless Communications and Mobile Computing

View full text Add to dashboard Cite

The introduction of deep transfer learning (DTL) further reduces the requirement of data and expert knowledge in various uses of applications, helping DNN-based models effectively reuse information. However, it often transfers all parameters from the source network that might be useful to the task. The redundant trainable parameters restrict DTL in low-computing-power devices and edge computing, while small effective networks with fewer parameters have difficulty transferring knowledge due to structural differences in design. For the challenge of how to transfer a simplified model from a complex network, in this paper, an algorithm is proposed to realize a sparse DTL, which only transfers and retains the most necessary structure to reduce the parameters of the final model. Sparse transfer hypothesis is introduced, in which a compressing strategy is designed to construct deep sparse networks that distill useful information in the auxiliary domain, improving the transfer efficiency. The proposed method is evaluated on representative datasets and applied for smart agriculture to train deep identification models that can effectively detect new pests using few data samples.

show abstract

“…This paper proposes the use of the improved predictive sparse decomposition (PSD) approach with DenseNet to remedy the above-mentioned problems. DenseNets have been successfully applied in medical image processing and classification [26][27][28]. Other applications of neural networks in medical image processing can also be found in [30,31,34].…”

Section: Related Workmentioning

confidence: 99%

Improved Predictive Sparse Decomposition Method With Densenet for Prediction of Lung Cancer

Mienye¹,

Sun²,

Wang³

2020

IJC

View full text Add to dashboard Cite

Lung cancer is the second most common form of cancer in both men and women. It is responsible for at least 25% of all cancer-related deaths in the United States alone. Accurate and early diagnosis of this form of cancer can increase the rate of survival. Computed tomography (CT) imaging is one of the most accurate techniques for diagnosing the disease. In order to improve the classification accuracy of pulmonary lesions indicating lung cancer, this paper presents an improved method for training a densely connected convolutional network (DenseNet). The optimized setting ensures that code prediction error and reconstruction error within hidden layers are simultaneously minimized. To achieve this and improve the classification accuracy of the DenseNet, we propose an improved predictive sparse decomposition (PSD) approach for extracting sparse features from the medical images. The sparse decomposition is achieved by using a linear combination of basis functions over the L2 norm. The effect of dropout and hidden layer expansion on the classification accuracy of the DenseNet is also investigated. CT scans of human lung samples are obtained from The Cancer Imaging Archive (TCIA) hosted by the University of Arkansas for Medical Sciences (UAMS). The proposed method outperforms seven other neural network architectures and machine learning algorithms with a classification accuracy of 95%.

show abstract

Multiple Feature Reweight DenseNet for Image Classification

Cited by 176 publications

References 23 publications

Identification of crystal symmetry from noisy diffraction patterns by a shape analysis and deep learning

Identification of crystal symmetry from noisy diffraction patterns by a shape analysis and deep learning

A Sparse Deep Transfer Learning Model and Its Application for Smart Agriculture

Improved Predictive Sparse Decomposition Method With Densenet for Prediction of Lung Cancer

Contact Info

Product

Resources

About