Editorial

3 pages, 156 KiB

Open AccessEditorial

Editorial for the Special Issue on Advanced Machine Learning Techniques for Sensing and Imaging Applications

by Bihan Wen and Zhangyang Wang

Micromachines 2022, 13(7), 1030; https://doi.org/10.3390/mi13071030 - 29 Jun 2022

Viewed by 994

Recent advances in machine learning, from large-scale optimization to building deep neural networks, are increasingly being applied in the emerging field of computational sensing and imaging [...] Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

Research

Jump to: Editorial

16 pages, 1858 KiB

Open AccessArticle

Towards More Efficient Security Inspection via Deep Learning: A Task-Driven X-ray Image Cropping Scheme

by Hong Duc Nguyen, Rizhao Cai, Heng Zhao, Alex C. Kot and Bihan Wen

Micromachines 2022, 13(4), 565; https://doi.org/10.3390/mi13040565 - 31 Mar 2022

Cited by 15 | Viewed by 3193

Abstract

X-ray imaging machines are widely used in border control checkpoints or public transportation, for luggage scanning and inspection. Recent advances in deep learning enabled automatic object detection of X-ray imaging results to largely reduce labor costs. Compared to tasks on natural images, object [...] Read more.

X-ray imaging machines are widely used in border control checkpoints or public transportation, for luggage scanning and inspection. Recent advances in deep learning enabled automatic object detection of X-ray imaging results to largely reduce labor costs. Compared to tasks on natural images, object detection for X-ray inspection are typically more challenging, due to the varied sizes and aspect ratios of X-ray images, random locations of the small target objects within the redundant background region, etc. In practice, we show that directly applying off-the-shelf deep learning-based detection algorithms for X-ray imagery can be highly time-consuming and ineffective. To this end, we propose a Task-Driven Cropping scheme, dubbed TDC, for improving the deep image detection algorithms towards efficient and effective luggage inspection via X-ray images. Instead of processing the whole X-ray images for object detection, we propose a two-stage strategy, which first adaptively crops X-ray images and only preserves the task-related regions, i.e., the luggage regions for security inspection. A task-specific deep feature extractor is used to rapidly identify the importance of each X-ray image pixel. Only the regions that are useful and related to the detection tasks are kept and passed to the follow-up deep detector. The varied-scale X-ray images are thus reduced to the same size and aspect ratio, which enables a more efficient deep detection pipeline. Besides, to benchmark the effectiveness of X-ray image detection algorithms, we propose a novel dataset for X-ray image detection, dubbed SIXray-D, based on the popular SIXray dataset. In SIXray-D, we provide the complete and more accurate annotations of both object classes and bounding boxes, which enables model training for supervised X-ray detection methods. Our results show that our proposed TDC algorithm can effectively boost popular detection algorithms, by achieving better detection mAPs or reducing the run time. Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

► Show Figures

Figure 1

14 pages, 6841 KiB

Open AccessArticle

Lightweight Multi-Scale Asymmetric Attention Network for Image Super-Resolution

by Min Zhang, Huibin Wang, Zhen Zhang, Zhe Chen and Jie Shen

Micromachines 2022, 13(1), 54; https://doi.org/10.3390/mi13010054 - 29 Dec 2021

Cited by 4 | Viewed by 1767

Abstract

Recently, with the development of convolutional neural networks, single-image super-resolution (SISR) has achieved better performance. However, the practical application of image super-resolution is limited by a large number of parameters and calculations. In this work, we present a lightweight multi-scale asymmetric attention network [...] Read more.

Recently, with the development of convolutional neural networks, single-image super-resolution (SISR) has achieved better performance. However, the practical application of image super-resolution is limited by a large number of parameters and calculations. In this work, we present a lightweight multi-scale asymmetric attention network (MAAN), which consists of a coarse-grained feature block (CFB), fine-grained feature blocks (FFBs), and a reconstruction block (RB). MAAN adopts multiple paths to facilitate information flow and accomplish a better balance of performance and parameters. Specifically, the FFB applies a multi-scale attention residual block (MARB) to capture richer features by exploiting the pixel-to-pixel correlation feature. The asymmetric multi-weights attention blocks (AMABs) in MARB are designed to obtain the attention maps for improving SISR efficiency and readiness. Extensive experimental results show that our method has comparable performance with fewer parameters than the current advanced lightweight SISR. Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

► Show Figures

Figure 1

25 pages, 2183 KiB

Open AccessFeature PaperEditor’s ChoiceArticle

Synergy Factorized Bilinear Network with a Dual Suppression Strategy for Brain Tumor Classification in MRI

by Guanghua Xiao, Huibin Wang, Jie Shen, Zhe Chen, Zhen Zhang and Xiaomin Ge

Micromachines 2022, 13(1), 15; https://doi.org/10.3390/mi13010015 - 23 Dec 2021

Cited by 7 | Viewed by 2247

Abstract

Automatic brain tumor classification is a practicable means of accelerating clinical diagnosis. Recently, deep convolutional neural network (CNN) training with MRI datasets has succeeded in computer-aided diagnostic (CAD) systems. To further improve the classification performance of CNNs, there is still a difficult path [...] Read more.

Automatic brain tumor classification is a practicable means of accelerating clinical diagnosis. Recently, deep convolutional neural network (CNN) training with MRI datasets has succeeded in computer-aided diagnostic (CAD) systems. To further improve the classification performance of CNNs, there is still a difficult path forward with regards to subtle discriminative details among brain tumors. We note that the existing methods heavily rely on data-driven convolutional models while overlooking what makes a class different from the others. Our study proposes to guide the network to find exact differences among similar tumor classes. We first present a “dual suppression encoding” block tailored to brain tumor MRIs, which diverges two paths from our network to refine global orderless information and local spatial representations. The aim is to use more valuable clues for correct classes by reducing the impact of negative global features and extending the attention of salient local parts. Then we introduce a “factorized bilinear encoding” layer for feature fusion. The aim is to generate compact and discriminative representations. Finally, the synergy between these two components forms a pipeline that learns in an end-to-end way. Extensive experiments exhibited superior classification performance in qualitative and quantitative evaluation on three datasets. Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

► Show Figures

Figure 1

11 pages, 1878 KiB

Open AccessArticle

Three-Dimensional Semantic Segmentation of Pituitary Adenomas Based on the Deep Learning Framework-nnU-Net: A Clinical Perspective

by Xujun Shu, Yijie Zhou, Fangye Li, Tao Zhou, Xianghui Meng, Fuyu Wang, Zhizhong Zhang, Jian Pu and Bainan Xu

Micromachines 2021, 12(12), 1473; https://doi.org/10.3390/mi12121473 - 29 Nov 2021

Cited by 9 | Viewed by 2209

Abstract

This study developed and evaluated nnU-Net models for three-dimensional semantic segmentation of pituitary adenomas (PAs) from contrast-enhanced T1 (T1ce) images, with aims to train a deep learning-based model cost-effectively and apply it to clinical practice. Methods: This study was conducted in two phases. [...] Read more.

This study developed and evaluated nnU-Net models for three-dimensional semantic segmentation of pituitary adenomas (PAs) from contrast-enhanced T1 (T1ce) images, with aims to train a deep learning-based model cost-effectively and apply it to clinical practice. Methods: This study was conducted in two phases. In phase one, two models were trained with nnUNet using distinct PA datasets. Model 1 was trained with 208 PAs in total, and model 2 was trained with 109 primary nonfunctional pituitary adenomas (NFPA). In phase two, the performances of the two models were investigated according to the Dice similarity coefficient (DSC) in the leave-out test dataset. Results: Both models performed well (DSC > 0.8) for PAs with volumes > 1000 mm³, but unsatisfactorily (DSC < 0.5) for PAs < 1000 mm³. Conclusions: Both nnU-Net models showed good segmentation performance for PAs > 1000 mm³ (75% of the dataset) and limited performance for PAs < 1000 mm³ (25% of the dataset). Model 2 trained with fewer samples was more cost-effective. We propose to combine the use of model-based segmentation for PA > 1000 mm³ and manual segmentation for PA < 1000 mm³ in clinical practice at the current stage. Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

► Show Figures

Figure 1

18 pages, 3650 KiB

Open AccessArticle

Wavelet Frequency Separation Attention Network for Chest X-ray Image Super-Resolution

by Yue Yu, Kun She and Jinhua Liu

Micromachines 2021, 12(11), 1418; https://doi.org/10.3390/mi12111418 - 18 Nov 2021

Cited by 6 | Viewed by 2061

Abstract

Medical imaging is widely used in medical diagnosis. The low-resolution image caused by high hardware cost and poor imaging technology leads to the loss of relevant features and even fine texture. Obtaining high-quality medical images plays an important role in disease diagnosis. A [...] Read more.

Medical imaging is widely used in medical diagnosis. The low-resolution image caused by high hardware cost and poor imaging technology leads to the loss of relevant features and even fine texture. Obtaining high-quality medical images plays an important role in disease diagnosis. A surge of deep learning approaches has recently demonstrated high-quality reconstruction for medical image super-resolution. In this work, we propose a light-weight wavelet frequency separation attention network for medical image super-resolution (WFSAN). WFSAN is designed with separated-path for wavelet sub-bands to predict the wavelet coefficients, considering that image data characteristics are different in the wavelet domain and spatial domain. In addition, different activation functions are selected to fit the coefficients. Inputs comprise approximate sub-bands and detail sub-bands of low-resolution wavelet coefficients. In the separated-path network, detail sub-bands, which have more sparsity, are trained to enhance high frequency information. An attention extension ghost block is designed to generate the features more efficiently. All results obtained from fusing layers are contracted to reconstruct the approximate and detail wavelet coefficients of the high-resolution image. In the end, the super-resolution results are generated by inverse wavelet transform. Experimental results show that WFSAN has competitive performance against state-of-the-art lightweight medical imaging methods in terms of quality and quantitative metrics. Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

► Show Figures

Figure 1

16 pages, 4973 KiB

Open AccessArticle

A 3D-2D Multibranch Feature Fusion and Dense Attention Network for Hyperspectral Image Classification

by Hongmin Gao, Yiyan Zhang, Yunfei Zhang, Zhonghao Chen, Chenming Li and Hui Zhou

Micromachines 2021, 12(10), 1271; https://doi.org/10.3390/mi12101271 - 18 Oct 2021

Cited by 2 | Viewed by 1649

Abstract

In recent years, hyperspectral image classification (HSI) has attracted considerable attention. Various methods based on convolution neural networks have achieved outstanding classification results. However, most of them exited the defects of underutilization of spectral-spatial features, redundant information, and convergence difficulty. To address these [...] Read more.

In recent years, hyperspectral image classification (HSI) has attracted considerable attention. Various methods based on convolution neural networks have achieved outstanding classification results. However, most of them exited the defects of underutilization of spectral-spatial features, redundant information, and convergence difficulty. To address these problems, a novel 3D-2D multibranch feature fusion and dense attention network are proposed for HSI classification. Specifically, the 3D multibranch feature fusion module integrates multiple receptive fields in spatial and spectral dimensions to obtain shallow features. Then, a 2D densely connected attention module consists of densely connected layers and spatial-channel attention block. The former is used to alleviate the gradient vanishing and enhance the feature reuse during the training process. The latter emphasizes meaningful features and suppresses the interfering information along the two principal dimensions: channel and spatial axes. The experimental results on four benchmark hyperspectral images datasets demonstrate that the model can effectively improve the classification performance with great robustness. Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

► Show Figures

Figure 1

21 pages, 2376 KiB

Open AccessArticle

Simultaneous Patch-Group Sparse Coding with Dual-Weighted ℓ_p Minimization for Image Restoration

by Jiachao Zhang, Ying Tong and Liangbao Jiao

Micromachines 2021, 12(10), 1205; https://doi.org/10.3390/mi12101205 - 01 Oct 2021

Cited by 2 | Viewed by 1785

Abstract

Sparse coding (SC) models have been proven as powerful tools applied in image restoration tasks, such as patch sparse coding (PSC) and group sparse coding (GSC). However, these two kinds of SC models have their respective drawbacks. PSC tends to generate visually annoying [...] Read more.

Sparse coding (SC) models have been proven as powerful tools applied in image restoration tasks, such as patch sparse coding (PSC) and group sparse coding (GSC). However, these two kinds of SC models have their respective drawbacks. PSC tends to generate visually annoying blocking artifacts, while GSC models usually produce over-smooth effects. Moreover, conventional

ℓ_{1}

minimization-based convex regularization was usually employed as a standard scheme for estimating sparse signals, but it cannot achieve an accurate sparse solution under many realistic situations. In this paper, we propose a novel approach for image restoration via simultaneous patch-group sparse coding (SPG-SC) with dual-weighted

ℓ_{p}

minimization. Specifically, in contrast to existing SC-based methods, the proposed SPG-SC conducts the local sparsity and nonlocal sparse representation simultaneously. A dual-weighted

ℓ_{p}

minimization-based non-convex regularization is proposed to improve the sparse representation capability of the proposed SPG-SC. To make the optimization tractable, a non-convex generalized iteration shrinkage algorithm based on the alternating direction method of multipliers (ADMM) framework is developed to solve the proposed SPG-SC model. Extensive experimental results on two image restoration tasks, including image inpainting and image deblurring, demonstrate that the proposed SPG-SC outperforms many state-of-the-art algorithms in terms of both objective and perceptual quality. Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

► Show Figures

Figure 1

15 pages, 1804 KiB

Open AccessArticle

Extreme Low-Resolution Activity Recognition Using a Super-Resolution-Oriented Generative Adversarial Network

by Mingzheng Hou, Song Liu, Jiliu Zhou, Yi Zhang and Ziliang Feng

Micromachines 2021, 12(6), 670; https://doi.org/10.3390/mi12060670 - 08 Jun 2021

Cited by 8 | Viewed by 2300

Abstract

Activity recognition is a fundamental and crucial task in computer vision. Impressive results have been achieved for activity recognition in high-resolution videos, but for extreme low-resolution videos, which capture the action information at a distance and are vital for preserving privacy, the performance [...] Read more.

Activity recognition is a fundamental and crucial task in computer vision. Impressive results have been achieved for activity recognition in high-resolution videos, but for extreme low-resolution videos, which capture the action information at a distance and are vital for preserving privacy, the performance of activity recognition algorithms is far from satisfactory. The reason is that extreme low-resolution (e.g., 12 × 16 pixels) images lack adequate scene and appearance information, which is needed for efficient recognition. To address this problem, we propose a super-resolution-driven generative adversarial network for activity recognition. To fully take advantage of the latent information in low-resolution images, a powerful network module is employed to super-resolve the extremely low-resolution images with a large scale factor. Then, a general activity recognition network is applied to analyze the super-resolved video clips. Extensive experiments on two public benchmarks were conducted to evaluate the effectiveness of our proposed method. The results demonstrate that our method outperforms several state-of-the-art low-resolution activity recognition approaches. Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

► Show Figures

Figure 1

13 pages, 2386 KiB

Open AccessArticle

Hybrid Dilated Convolution with Multi-Scale Residual Fusion Network for Hyperspectral Image Classification

by Chenming Li, Zelin Qiu, Xueying Cao, Zhonghao Chen, Hongmin Gao and Zaijun Hua

Micromachines 2021, 12(5), 545; https://doi.org/10.3390/mi12050545 - 10 May 2021

Cited by 21 | Viewed by 3096

Abstract

The convolutional neural network (CNN) has been proven to have better performance in hyperspectral image (HSI) classification than traditional methods. Traditional CNN on hyperspectral image classification is used to pay more attention to spectral features and ignore spatial information. In this paper, a [...] Read more.

The convolutional neural network (CNN) has been proven to have better performance in hyperspectral image (HSI) classification than traditional methods. Traditional CNN on hyperspectral image classification is used to pay more attention to spectral features and ignore spatial information. In this paper, a new HSI model called local and hybrid dilated convolution fusion network (LDFN) was proposed, which fuses the local information of details and rich spatial features by expanding the perception field. The details of our local and hybrid dilated convolution fusion network methods are as follows. First, many operations are selected, such as standard convolution, average pooling, dropout and batch normalization. Then, fusion operations of local and hybrid dilated convolution are included to extract rich spatial-spectral information. Last, different convolution layers are gathered into residual fusion networks and finally input into the softmax layer to classify. Three widely hyperspectral datasets (i.e., Salinas, Pavia University and Indian Pines) have been used in the experiments, which show that LDFN outperforms state-of-art classifiers. Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

► Show Figures

Figure 1

13 pages, 4443 KiB

Open AccessFeature PaperEditor’s ChoiceArticle

Machine Learning-Based Pipeline for High Accuracy Bioparticle Sizing

by Shaobo Luo, Yi Zhang, Kim Truc Nguyen, Shilun Feng, Yuzhi Shi, Yang Liu, Paul Hutchinson, Giovanni Chierchia, Hugues Talbot, Tarik Bourouina, Xudong Jiang and Ai Qun Liu

Micromachines 2020, 11(12), 1084; https://doi.org/10.3390/mi11121084 - 07 Dec 2020

Cited by 5 | Viewed by 2822

Abstract

High accuracy measurement of size is essential in physical and biomedical sciences. Various sizing techniques have been widely used in sorting colloidal materials, analyzing bioparticles and monitoring the qualities of food and atmosphere. Most imaging-free methods such as light scattering measure the averaged [...] Read more.

High accuracy measurement of size is essential in physical and biomedical sciences. Various sizing techniques have been widely used in sorting colloidal materials, analyzing bioparticles and monitoring the qualities of food and atmosphere. Most imaging-free methods such as light scattering measure the averaged size of particles and have difficulties in determining non-spherical particles. Imaging acquisition using camera is capable of observing individual nanoparticles in real time, but the accuracy is compromised by the image defocusing and instrumental calibration. In this work, a machine learning-based pipeline is developed to facilitate a high accuracy imaging-based particle sizing. The pipeline consists of an image segmentation module for cell identification and a machine learning model for accurate pixel-to-size conversion. The results manifest a significantly improved accuracy, showing great potential for a wide range of applications in environmental sensing, biomedical diagnostical, and material characterization. Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

► Show Figures

Figure 1

16 pages, 2111 KiB

Open AccessArticle

Object Detection at Level Crossing Using Deep Learning

by Muhammad Asad Bilal Fayyaz and Christopher Johnson

Micromachines 2020, 11(12), 1055; https://doi.org/10.3390/mi11121055 - 29 Nov 2020

Cited by 23 | Viewed by 3910

Abstract

Multiple projects within the rail industry across different regions have been initiated to address the issue of over-population. These expansion plans and upgrade of technologies increases the number of intersections, junctions, and level crossings. A level crossing is where a railway line is [...] Read more.

Multiple projects within the rail industry across different regions have been initiated to address the issue of over-population. These expansion plans and upgrade of technologies increases the number of intersections, junctions, and level crossings. A level crossing is where a railway line is crossed by a road or right of way on the level without the use of a tunnel or bridge. Level crossings still pose a significant risk to the public, which often leads to serious accidents between rail, road, and footpath users and the risk is dependent on their unpredictable behavior. For Great Britain, there were three fatalities and 385 near misses at level crossings in 2015–2016. Furthermore, in its annual safety report, the Rail Safety and Standards Board (RSSB) highlighted the risk of incidents at level crossings during 2016/17 with a further six fatalities at level crossings including four pedestrians and two road vehicles. The relevant authorities have suggested an upgrade of the existing sensing system and the integration of new novel technology at level crossings. The present work addresses this key issue and discusses the current sensing systems along with the relevant algorithms used for post-processing the information. The given information is adequate for a manual operator to make a decision or start an automated operational cycle. Traditional sensors have certain limitations and are often installed as a “single sensor”. The single sensor does not provide sufficient information; hence another sensor is required. The algorithms integrated with these sensing systems rely on the traditional approach, where background pixels are compared with new pixels. Such an approach is not effective in a dynamic and complex environment. The proposed model integrates deep learning technology with the current Vision system (e.g., CCTV to detect and localize an object at a level crossing). The proposed sensing system should be able to detect and localize particular objects (e.g., pedestrians, bicycles, and vehicles at level crossing areas.) The radar system is also discussed for a “two out of two” logic interlocking system in case of fail-mechanism. Different techniques to train a deep learning model are discussed along with their respective results. The model achieved an accuracy of about 88% from the MobileNet model for classification and a loss metric of 0.092 for object detection. Some related future work is also discussed. Full article

(This article belongs to the Special Issue Advanced Machine Learning Techniques for Sensing and Imaging Applications)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Advanced Machine Learning Techniques for Sensing and Imaging Applications

Share This Special Issue

Special Issue Editors

Special Issue Information

Related Special Issue

Published Papers (12 papers)

Editorial

Research

Further Information

Guidelines

MDPI Initiatives

Follow MDPI