Search

In the field of image object detection and semantic segmentation, improving the accuracy of object identification and segmentation is a primary goal. To achieve this, leveraging the potential of multi-scale information through feature map refinement and fusion has been widely recognized. However, existing feature fusion methods either design more complex feature pyramid networks, replace existing detectors, or incrementally introduce feature fusion modules, overlooking the effective approach of enhancing spatial information in deep feature maps. We propose a novel pluggable feature fusion paradigm termed 'Effective Learning Bridge'. Our research introduces an efficient and adaptive learning mechanism that builds learning bridges between feature maps at different scales within the feature pyramid, thereby enhancing the spatial information of objects in deep feature maps. This mechanism is specifically designed for multi-scale feature maps and can be seamlessly integrated into any network incorporating feature maps. By altering the model's backpropagation path, we successfully improve learning efficiency, which in turn enhances the accuracy of object detection and segmentation. Our proposed paradigm and method were extensively evaluated through experiments on SIMD, HRSID, and WHDLD datasets and benchmark models. The results unequivocally demonstrate the effectiveness of our approach in significantly improving the accuracy of object detection and semantic segmentation, as well as the overall learning efficiency of the model. [ABSTRACT FROM AUTHOR]

This paper proposes a channel pruning method based on multi-stage feature de-correlation to obtain a more efficient convolutional neural network (CNN) model. Based on the correlation of hidden features at each level of the network, we refine more efficient features of each convolutional layer by applying feature de-correlation constraints (MFD Loss) to each convolutional layer of the network and then prune channels according to the modulus of the feature maps output from each layer. After several rounds of pruning and fine-tuning, a network with similar accuracy, a substantially smaller network size, and more efficient operation is generated compared to the original model. Our experiments on pruning various popular CNN models on many standard datasets demonstrate the method’s effectiveness. Specifically, for VGG-16 on CIFAR10, our approach eliminates parameters by 97.0%, saves Float-Point-Operations (FLOPs) by 66.9%, with a 0.4% accuracy gain and state-of-art performance. For ResNet-50 on the ImageNet dataset, our method eliminates parameters by 30.0%, and saves FLOPs by 52%, with 1.4% accuracy loss, which also proves the effectiveness of the method. The code for the paper can be found at https://github.com/lovelyemperor/MFD. [ABSTRACT FROM AUTHOR]

Diabetes is a common and serious global disease that damages blood vessels in the eye, leading to vision loss. Early and accurate diagnosis of this issue is crucial to reduce the risk of visual impairment. The typical deep learning (DL) methods for diabetic retinopathy (DR) grading are often time‐consuming, resulting in unsatisfactory detection performance due to inadequate representation of lesion features. To overcome these challenges, this research proposes a new automated mechanism for detecting and classifying DR, aiming to identify DR severities and different stages. To figure out and capture feature characteristics from DR samples, a conjugated attention mechanism and vision transformer are utilized within a collective net model, which automatically generates feature maps for diagnosing DR. These extracted feature maps are then fused through the feature fusion function in a fused attention net model, calculating attention weights to produce the most powerful feature map. Finally, the DR cases are identified and discriminated using the kernel extreme learning machine (KELM) model. For evaluating DR severity, our work utilizes four different benchmark datasets: APTOS 2019, MESSIDOR‐2 dataset, DiaRetDB1 V2.1, and DIARETDB0 datasets. To illuminate data noise and unwanted variations, two preprocessing steps are carried out, which include contrast enhancement and illumination correction. The experimental results evaluated using well‐known indicators demonstrate that the suggested method achieves a higher accuracy of 99.63% compared to other baseline methods. This research contributes to the development of powerful DR screening techniques that are less time‐consuming and capable of automatically identifying DR severity levels at a premature level. [ABSTRACT FROM AUTHOR]

Maintaining a high input resolution is crucial for more complex tasks like detection or segmentation to ensure that models can adequately identify and reflect fine details in the output. This study aims to reduce the computation costs associated with high-resolution input by using a variant of transformer, known as the Adaptive Clustering Transformer (ACT). The proposed model is named ACT-FRCNN. Which integrates ACT with a Faster Region-Based Convolution Neural Network (FRCNN) for a detection task head. In this paper, we proposed a method to improve the detection framework, resulting in better performance for out-of-domain images, improved object identification, and reduced dependence on non-maximum suppression. The ACT-FRCNN represents a significant step in the application of transformer models to challenging visual tasks like object detection, laying the foundation for future work using transformer models. The performance of ACT-FRCNN was evaluated on a variety of well-known datasets including BSDS500, NYUDv2, and COCO. The results indicate that ACT-FRCNN reduces over-detection errors and improves the detection of large objects. The findings from this research have practical implications for object detection and other computer vision tasks. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Ordnance Equipment Engineering is the property of Chongqing University of Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

We introduce a new embedding technique based on a nested barycentric coordinate system. We show that our embedding can be used to transform the problems of polyhedron approximation, piecewise linear classification and convex regression into one of finding a linear classifier or regressor in a higher dimensional (but nevertheless quite sparse) representation. Our embedding maps a piecewise linear function into an everywhere-linear function, and allows us to invoke well-known algorithms for the latter problem to solve the former. We explain the applications of our embedding to the problems of approximating separating polyhedra—in fact, it can approximate any convex body and unions of convex bodies—as well as to classification by separating polyhedra, and to piecewise linear regression. [ABSTRACT FROM AUTHOR]

Nowadays, Convolutional neural networks (CNNs) have achieved relevant results on several data sciences-related tasks, such as image processing and pattern recognition. However, CNNs contain an immense number of parameters which often leads to a huge redundancy, overfitting, and a significant amount of memory. In this paper, we aim to present a multi-objective optimization model for kernels redundancy reduction in convolutional neural networks. In fact, the suggested approach, named MOFM-CNN, allows to minimize redundant feature maps using a set of decision control variables. MOFM-CNN is composed of two objectives where in the first one, the decision variables are technically introduced in the cross-entropy function in order to evaluate the impact of each feature map on the CNNs training. In the second one, the control parameters are used to calculate the proportion of active feature maps, that is related to the complexity of the model. The resultant problem is manipulated and solved using non dominated sorting genetic algorithm (NSGA-II). The performance of our proposal is demonstrated visually and numerically for both classification and features maps optimization. [ABSTRACT FROM AUTHOR]

The aim of this paper is to develop a marginal screening method for variable screening in high-dimensional binary classification based on the Wasserstein distance accounting for the distributional difference. Many existing screening methods, such as the two-sample t-test and Kolmogorov test, have been developed under the parametric/nonparametric modeling assumptions to reduce the dimension of the predictors. However, such modeling specifications or nonparametric approaches are associated with the probability measure induced by the predictor in a Euclidean space. While many machine learning methods have successfully found the nonlinear decision boundary in the transformed space, called the reproducing kernel Hilbert space (RKHS), we consider the Wasserstein filter's capacity to detect the distributional difference between two probability measures induced by the nonlinear function of the predictor in the RKHS. Thereby, we can flexibly filter out the non-informative predictors associated with the binary classification, as well as escape the modeling assumptions required in a Euclidean space. We prove that the Wasserstein filter satisfies the sure screening property under some mild conditions. We also demonstrate the advantages of our proposed approach by comparing the finite sample performance of it with those of the existing choices through simulation studies, as well as through application to lung cancer data. [ABSTRACT FROM AUTHOR]

Channel pruning can reduce the number of neural network parameters and computational cost by eliminating redundant channels, its main purpose is to adapt to resource constrained devices. Evaluation-based global pruning and network search-based pruning are two common methods of channel pruning. However, the network architecture pruned by the global mask is often not optimal, while the method that directly searches for the optimal architecture will introduce a large number of hyperparameters, which greatly increases the training cost. In this paper, we propose a novel Two-dimensional information Entropy based Channel Pruning method (TECP). The pruning process consists of two steps. First, a global mask pruning scheme is employed to obtained a pre-pruning model. Then, the two-dimensional information entropy is calculated by using feature maps of dense network to adjust the pre-pruning model adaptively to get a compact network. Moreover, the entropy values are used to determine the minimum number of reserved channels per layer based on to avoid the imbalance of network architecture and the layer collapse caused by global pruning. Extensive experiments with a variety of networks on several datasets clearly demonstrate the effectiveness of our proposed TECP method. For example, results show that on CIFAR-10, the compressed model achieves comparable accuracy to the original model, but with a significantly lower number of parameters (44.29% for ResNet-20 and 46.79% for VGG-16). This is beneficial for industrial deployment. And experimental results also show that TECP method obtain the better performance compared with state-of-the-art method. [ABSTRACT FROM AUTHOR]

Synthetic aperture radar (SAR) has developed in leaps and bounds over the past decades, which makes rapid revisit and high-frequency coverage feasible. However, accurate and efficient registration of the SAR image is still a challenging task. Many existing SAR image registration methods major in describing detected features in a unique, identifiable, but maybe complex way. These descriptors are usually high-dimensional, resulting in increased computational complexity. To this end, a low-dimensional multiscale fast method for SAR image registration is proposed in this paper. First, the candidate points are detected in the modulus map of phase congruency (PC). This operation is robust to speckle noise in SAR images and improves the repeatability of feature points. Second, circular neighbourhoods of each point are extracted in multiple scales to describe their features with the maximum index map (MIM). Note that we condense the feature information of candidate points in the whole neighbourhood in an intensity-order way, which significantly reduces the dimensionality of the descriptors. Overall, the proposed method focuses on efficient representation of point features, thus allowing more feature points to be detected and involved in the subsequent high-speed feature matching. Experiments on raw SAR images with neither prior information nor any pre-processing measure like terrain correction and de-speckling demonstrate the efficacy of the proposed method over other state-of-the-art SAR image registration algorithms. Some analyses concerning the factors affecting feature matching and invariance of PC map and MIM are also studied.

With the continuous advancement of the economy and technology, the number of cars continues to increase, and the traffic congestion problem on some key roads is becoming increasingly serious. This paper proposes a new vehicle information feature map (VIFM) method and a multi-branch convolutional neural network (MBCNN) model and applies it to the problem of traffic congestion detection based on camera image data. The aim of this study is to build a deep learning model with traffic images as input and congestion detection results as output. It aims to provide a new method for automatic detection of traffic congestion. The deep learning-based method in this article can effectively utilize the existing massive camera network in the transportation system without requiring too much investment in hardware. This study first uses an object detection model to identify vehicles in images. Then, a method for extracting a VIFM is proposed. Finally, a traffic congestion detection model based on MBCNN is constructed. This paper verifies the application effect of this method in the Chinese City Traffic Image Database (CCTRIB). Compared to other convolutional neural networks, other deep learning models, and baseline models, the method proposed in this paper yields superior results. The method in this article obtained an F1 score of 98.61% and an accuracy of 98.62%. Experimental results show that this method effectively solves the problem of traffic congestion detection and provides a powerful tool for traffic management. [ABSTRACT FROM AUTHOR]

The current object detection algorithm suffers from slow detection speed and low object recognition rate when detecting marine benthos due to small objects, severe densities and occlusions. To solve these problems, an improved object detection algorithm based on FCOS, DSE-FCOS (Dilated & SE block reinforced FCOS), is proposed for marine benthos in this paper. In our method, to meet the needs of real-time detection, we only use a single-level feature map for detection, we propose DSE module to solve the problem of insufficient single-scale detection accuracy. The DW (Dual Weighting) label assignment method is introduced and optimized to make it more suitable for the detection of marine benthos, increasing the speed of detection while ensuring accuracy. We performed experiments on the DUO dataset. The results show that our method can improve mAP (mean average precision) by 5.9% on the basis of 13.8% speed improvement compared with FCOS. This improvement reflects that DSE-FCOS is more suitable for the detection of marine benthos. [ABSTRACT FROM AUTHOR]

Technology is always a double-edged sword, and with the astonishing advancements in technology, it is expected that the DeepFake problem will become more common and serious. DeepFake has recently caused a lot of trouble because its flaws outweigh its advantages. Since DeepFake has such a significant influence on individuals deception, instability of principles and falsification of evidence. Instead of just affecting people, it led to multiple incidents that affected the image of entire nations. In this paper, a model that has been built to mitigate the negative effects of deepFake and maintain an individual's reputation by detecting the alteration of people's photographs and videos. A model with integrated vision transformer architectures Deep-ViT and Cross-ViT is designed to process pre-extracted faces from FF++ dataset. The model distinguishes between the real and fake faces in two different perspectives, subclass detection on each manipulation method and overall detection of all types. The proposed model achieves an outstanding results and the highest accuracy in FaceSwap manipulation method with 98%. [ABSTRACT FROM AUTHOR]

Diabetic retinopathy (DR) involves retina damage due to diabetes, often leading to blindness. It is diagnosed via color fundus injections, but the manual analysis is cumbersome and error-prone. While computer vision techniques can predict DR stages, they are computationally intensive and struggle with complex data extraction. In this research, our prime objective was to automate the process of DR classification into its various stages using convolutional neural network (CNN) models. We employed the performance of fifteen pre-trained models with our novel proposed diabetic retinopathy network (DRNet13) model. We aimed to discern the most efficient model for accurate diabetic retinopathy (DR) staging based on fundus images from five DR classes. We preprocessed the image using a median filter for noise reduction and Gamma correction for image enhancement. We expanded our dataset from 3662 to 7500 images to create a more generalized training model through various augmentation techniques. We also evaluated multiple evaluation metrics, including accuracy, precision, F1-score, Sensitivity, Specificity, Area under the curve (AUC), Mean Squared Error (MSE), False Positive Rate (FPR), False Negative Rate (FNR), in addition to confusion matrices for an in-depth comparison of the performance of these models. Feature maps were employed to illuminate decision making areas in the DRNet13 model, which achieved a 97 % accuracy rate for DR detection, surpassing other CNN architectures in speed and efficiency. Despite a few misclassifications, the model's capability to identify critical features demonstrates its potential as an impactful diagnostic tool for timely and accurate identification of diabetic retinopathy.

In addressing the complex challenge of real-time and precise recognition of umpire signals in sporting events, we introduce ICCG-YOLO, a rapid and effective approach that builds upon the YOLO-v5 architecture. Our method innovatively incorporates Involution operations within the CSP components for superior spatial information modeling and channel-wise parameter sharing, significantly reducing parameter count while capturing extensive features through larger kernels. This stable approach between computational resource & performance and detection accuracy is further enhanced by integrating the CoordAttention block for precise localization and recognition of signals with minimal parameter addition. We also refine the model’s up-sampling process with the CARAFE (Content-Aware ReAssembly of FEatures) block, enabling content-driven feature enlargement that amplifies the receptive field without compromising the model’s compact stature. Complementing this, model compression is achieved using Ghost convolution, capitalizing on simple linear transformations for feature generation, paired with an improved activation function to activate all neural network neurons fully. This results in a multi-scale detection capability while maintaining a moderate depth and reducing overall model complexity. Our experiments on a custom dataset for umpire signal detection and the Chalearn dataset for general gesture recognition in diverse scenarios have successfully validated the high accuracy, rapid processing, and versatility of ICCG-YOLO.

Chemo-informatics is a recent development in the field of chemistry. It has helped scientists develop various chemical structures using computers, leading to the solution of some intricate problems in chemistry, including in medical chemistry. Objective: The mapping of literature in chemo-informatics was performed to reveal publication trends, authorship trends, core journals, and collaborative authors at the national, international, and institutional levels. The annual growth rate revealed the growth pattern of publications. The data of collaboration was visualized using the VOS Viewer tool. Methodology: The methodology used for this study was citation analysis of the publications downloaded from the SCOPUS database for the years 2017 to 2022. A total of 1425 data items were downloaded based on the keywords used for searching. Out of these 1425 after cleaning the data, only 1,413 citations were used for further study. Out of these 1,413 citations, 13 citations were anonymous so they were also ignored for authorship analysis. Findings: The findings of the study were that the annual growth rate of publications as shown by the highest number of publications (276) was in 2021 and the lowest 195 (13.80%) in 2018. Articles (65.46%) were the preferred publications by the authors. Multiple author (5 or more) publications show the prevalence of teamwork, which was further substantiated by the Av. Collaboration Index of 4.48; Av. Degree of collaboration was .0926 and Av. collaboration coefficient was .0179. J. Medina-Franco of Universidad Nacional Autónoma de México was the most prolific author with 53 publications during the 5 years of the study period. The Journal of Chemical Information and Modeling headed the list of publications with 48 publications while the Journal of Chemical Education received the highest number of citations (1052). In the list of countries with maximum contribution, the USA headed the list with 327 publications receiving 7735 citations. The international link strength was 211, which iswas revealed by the International Cooperation Index (ICI) of 64.52. China was placed in second place in international collaboration. All these papers could be only b possible through the use of computers, as the 21st century has allowed the use of computers in general studies which have become precursors to develop various models and theories based on computers. The present study reveals the collaborative nature of research which is very much prevalent nowadays and which can also be seen in the visualization report. The study also reveals that authors are adopting team research and have a wider perspective of their research.

Compared with large-scale ships, small-scale ships occupy few pixels and have low contrast, so it poses a great challenge to detect multiscale ships in synthetic aperture radar (SAR) images. In order to improve the accuracy of multiscale ship detection in SAR images, this article designs a general multiscale pyramid attention module (MPAM), which is a plug-and-play lightweight module that can adapt to many ship detection networks. In the MPAM, a deep feature extraction submodule is first designed to use the multiscale pyramid structure to divide the feature map into different levels, extracting rich features with resolution and semantic information for multiscale ship detection. The channel multilayer attention fusion submodule and spatial multilayer attention fusion submodule are then designed to fuse the channel and spatial attention blocks on different level feature maps, which could better learn the dependent features from the channel and spatial dimensions, to enhance the feature representation. Finally, the fused feature map is input into the existing ship detection networks to obtain the detection result. Experiments on SAR datasets containing multiscale ships show that the effectiveness of the MPAM in improving the accuracy of the existing ship detection networks.

In order to utilize the advantage of the feature map in calculating efficiency and solve the problem that the traditional dynamic window approach is sensitive to global parameters, a path planning fusion algorithm based on feature map is proposed. A feature map expression applicable to path planning is given, and the detection of obstacles in the feature map is achieved by improving the calculation method of the distance between the robot and the obstacles. Combined with the basic principle of the Bug algorithm and the properties of line segment features, the searching and optimization algorithm is used to search the global feasible path first, and then the key nodes of the global optimal path are obtained by node optimization, and solutions are proposed for the problems of search direction selection at internal and external corner points and obstacle endpoint bypassing. To address the problem of high sensitivity of the traditional dynamic window approach to global parameters, the degree of influence of the parameters in the objective function on the planned path when the robot is at different positions is analyzed, and the original objective function is improved using the dynamic parameter approach. When the algorithms are fused, the calculation method of direction function in the objective function is improved in order to solve the problem that the robot may slow down in the intermediate nodes of the path. The simulation experiment verifies that the searching optimization algorithm is effective, the improved dynamic window approach reduces the sensitivity of parameters, and the fusion algorithm has a greater advantage in computational efficiency, with a maximum reduction of 79.27% and a minimum reduction of 43.16% in computational time consumption, and the robot moves more smoothly.

Modulation classification (MC) is a critical task in wireless communication systems, enabling the identification of the modulation class in the received signals. In this paper, we analyzed a novel multi-layer convolutional neural network (CNN) to extract hierarchical features directly from the raw baseband samples. Moreover, we compared the training and testing accuracy of the CNN model for various decimation rates, input sample size and the number of convolutional layers. The results showed that the three-layer CNN model provided better classification accuracy with less computation cost. Furthermore, we observed that the MC performance of the proposed CNN model was better than the other deep learning (DL) and cumulant-based models.

With the continuous development of 3D data acquisition technology in recent years, it is more and more convenient to obtain the point cloud data of large scenes, which contains a variety of rich information. How to effectively and accurately classify and segment point cloud data of large scenes has become a research hot-spot in the field of computer vision. In this paper, we study the method based on clustering, make full use of the spatial location and context information, and propose a new network framework, FeatureB2SENet. In the 2D and 3D projection feature calculation, we generate a 32 × 32 × 1 feature image for each point and input it into the convolution neural network to process the feature image. Finally, a comprehensive verification analysis is carried out on GML _ A, GML _ B and Vaihingen data sets, which proves that the classification effect is better. [ABSTRACT FROM AUTHOR]

The process of feature mapping is used to represent features on a map with their corresponding properties. The features are displayed visually and their associated information is made available. Various abstraction approaches, including alignment approaches, are introduced to reduce the computational complexity. However, previous works based on standard treatments of feature maps often suffer from the negative effects of noise and background details. To solve this problem, we present a simple and effective implementation approach for Filter Pruning via Attention Consistency (FPAC), which determines a revolutionary filter pruning mechanism. Feature maps that concentrate on a single layer are inconsistent, which can affect the spatial dimension. The feature map with minimum consistency is less significant and is experimentally demonstrated. This study presents a novel layer-wise pruning technique using the Aphid Ant Mutualism (AAM) algorithm, which considers the sensitivity of various convolutional network layers to model inference and sets the optimal pruning ratio. The accuracy of the compressed model is enhanced by eliminating high redundancy through pruning correlated filters. The performance of FPAC is confirmed through the Caltech 256 image dataset. With VGG-16 on the Caltech 256 image dataset, the classification accuracy is enhanced from 93.96 to 94.03%. With ResNet-50 on the Caltech 256 image dataset, 45% FLOPs are achieved with an accuracy loss of only 0.53%. [ABSTRACT FROM AUTHOR]

With the medical internet of things, many automated diagnostic models related to eye diseases are easier. The doctors could quickly contrast and compare retina fundus images. The retina image contains a lot of information in the image. The task of detecting diabetic macular edema from retinal images in the healthcare system is difficult because the details in these images are very small. This paper proposed the new model based on the medical internet of things for predicting diabetic macular edema in retina fundus images. The method called DMER (Diabetic Macular Edema in Retina fundus images) to detect diabetic macular edema in retina fundus images based on improving deep residual network being combined with feature pyramid network in the context of the medical internet of things. The DMER method includes the following stages: (i) ResNet101 improved combining with feature pyramid network is used to extract features of the image and obtain the map of these features; (ii) a region proposal network to look for potential anomalies; and (iii) the predicted bounding boxes against the true bounding box by the regression method to certify the capability of macular edema. The MESSIDOR and DIARETDB1 datasets are used for testing with evaluation criteria such as sensitivity, specificity, and accuracy. The accuracy of the DMER method is about 98.08% with MESSIDOR dataset and 98.92% with DIARETDB1 dataset. The results of the method DMER are better than those of the other methods up to the present time with the above datasets. [ABSTRACT FROM AUTHOR]

Automated pavement crack image segmentation presents a significant challenge due to the difficulty in detecting slender cracks on complex pavement backgrounds, as well as the significant impact of lighting conditions. In this paper, we propose a novel approach for automated pavement crack detection using a multi-scale feature fusion network based on the Transformer architecture, leveraging an encoding-decoding structure. In the encoding phase, the Transformer is leveraged as a substitute for the convolution operation, which utilizes global modeling to enhance feature extraction capabilities and address long-distance dependence. Then, dilated convolution is employed to increase the receptive field of the feature map while maintaining resolution, thereby further improving context information acquisition. In the decoding phase, the linear layer is employed to adjust the length of feature sequence output by different encoder block, and the multi-scale feature map is obtained after dimension conversion. Detailed information of cracks can be restored by fusing multi-scale features, thereby improving the accuracy of crack detection. Our proposed method achieves an F1 score of 70.84% on the Crack500 dataset and 84.50% on the DeepCrack dataset, which are improvements of 1.42% and 2.07% over the state-of-the-art method, respectively. The experimental results show that the proposed method has higher detection accuracy, better generalization and better crack detection results can be obtained under both high and low brightness conditions.

Copyright of Journal of Frontiers of Computer Science & Technology is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Lung-related diseases continue to be a leading cause of global mortality. Timely and precise diagnosis is crucial to save lives, but the availability of testing equipment remains a challenge, often coupled with issues of reliability. Recent research has highlighted the potential of Chest X-ray (CXR) images in identifying various lung diseases, including COVID-19, fibrosis, pneumonia, and more. In this comprehensive study, four publicly accessible datasets have been combined to create a robust dataset comprising 6650 CXR images, categorized into seven distinct disease groups. To effectively distinguish between normal and six different lung-related diseases (namely, bacterial pneumonia, COVID-19, fibrosis, lung opacity, tuberculosis, and viral pneumonia), a Deep Learning (DL) architecture called a Multi-Scale Convolutional Neural Network (MS-CNN) is introduced. The model is adapted to classify multiple numbers of lung disease classes, which is considered to be a persistent challenge in the field. While prior studies have demonstrated high accuracy in binary and limited-class scenarios, the proposed framework maintains this accuracy across a diverse range of lung conditions. The innovative model harnesses the power of combining predictions from multiple feature maps at different resolution scales, significantly enhancing disease classification accuracy. The approach aims to shorten testing duration compared to the state-of-the-art models, offering a potential solution toward expediting medical interventions for patients with lung-related diseases and integrating explainable AI (XAI) for enhancing prediction capability. The results demonstrated an impressive accuracy of 96.05%, with average values for precision, recall, F1-score, and AUC at 0.97, 0.95, 0.95, and 0.94, respectively, for the seven-class classification. The model exhibited exceptional performance across multi-class classifications, achieving accuracy rates of 100%, 99.65%, 99.21%, 98.67%, and 97.47% for two, three, four, five, and six-class scenarios, respectively. The novel approach not only surpasses many pre-existing state-of-the-art (SOTA) methodologies but also sets a new standard for the diagnosis of lung-affected diseases using multi-class CXR data. Furthermore, the integration of XAI techniques such as SHAP and Grad-CAM enhanced the transparency and interpretability of the model's predictions. The findings hold immense promise for accelerating and improving the accuracy and confidence of diagnostic decisions in the field of lung disease identification. [ABSTRACT FROM AUTHOR]

Modulation classification (MC) is a critical task in wireless communication systems, enabling the identification of the modulation class in the received signals. In this paper, we analyzed a novel multi-layer convolutional neural network (CNN) to extract hierarchical features directly from the raw baseband samples. Moreover, we compared the training and testing accuracy of the CNN model for various decimation rates, input sample size and the number of convolutional layers. The results showed that the three-layer CNN model provided better classification accuracy with less computation cost. Furthermore, we observed that the MC performance of the proposed CNN model was better than the other deep learning (DL) and cumulant-based models. [ABSTRACT FROM AUTHOR]

Single Shot Multibox Detector (SSD) uses multi-scale feature maps to detect and recognize objects, which considers the advantages of both accuracy and speed, but it is still limited to detecting small-sized objects. Many researchers design new detectors to improve the accuracy by changing the structure of the multi-scale feature pyramid which has proved very useful. But most of them only simply merge several feature maps without making full use of the close connection between features with different scales. In contrast, a novel feature fusion module and an effective feature enhancement module is proposed, which can significantly improve the performance of the original SSD. In the feature fusion module, the feature pyramid is produced through iteratively fusing three feature maps with different receptive fields to obtain contextual information. In the feature enhancement module, the features are enhanced along the channel and spatial dimensions at the same time to improve their expression ability. Our network can achieve 82.5% mean Average Precision (mAP) on the VOC 2007 t e s t , 81.4% mAP on the VOC 2012 t e s t and 34.8% mAP on COCO t e s t - d e v 2017, respectively, with the input size 5 1 2 × 5 1 2. Comparative experiments prove that our method outperforms many state-of-the-art detectors in both aspects of accuracy and speed. [ABSTRACT FROM AUTHOR]

Fine-Grained Visual Classification (FGVC) has consistently been challenging in various domains, such as aviation and animal breeds. It is mainly due to the FGVC’s criteria that differ with a considerably small range or subtle pattern differences. In the deep convolutional neural network, the covariance between feature maps positively affects the selection of features to learn discriminative regions automatically. In this study, we propose a method for a fine-grained classification model by inserting an attention module that uses covariance characteristics. Specifically, we introduce a feature map attention module (FCA) to extract the feature map between convolution blocks, constituting the existing classification model. The FCA module then applies the corresponding value of the covariance matrix to the channel to focus on the salient area. We demonstrate the need for fine-grained classification in a hierarchical manner by focusing on the diverse scale representation. Additionally, we implemented two ablation studies to show how each suggested strategy affects classification performance. Our experiments are conducted on three datasets, CUB-200-2011, Stanford Cars, and FGVC-Aircraft, primarily used for fine-grained classification tasks. Our method outperforms the state-of-the-art models by a margin of 0.4%, 1.1%, and 1.4%.

We propose an improved BM3D algorithm for block-matching based on UNet denoising network feature maps and structural similarity (SSIM). In response to the traditional BM3D algorithm that directly performs block-matching on a noisy image, without considering the deep-level features of the image, we propose a method that performs block-matching on the feature maps of the noisy image. In this method, we perform block-matching on multiple depth feature maps of a noisy image, and then determine the positions of the corresponding similar blocks in the noisy image based on the block-matching results, to obtain the set of similar blocks that take into account the deep-level features of the noisy image. In addition, we improve the similarity measure criterion for block-matching based on the Structural Similarity Index, which takes into account the pixel-by-pixel value differences in the image blocks while fully considering the structure, brightness, and contrast information of the image blocks. To verify the effectiveness of the proposed method, we conduct extensive comparative experiments. The experimental results demonstrate that the proposed method not only effectively enhances the denoising performance of the image, but also preserves the detailed features of the image and improves the visual quality of the denoised image. [ABSTRACT FROM AUTHOR]

Most models in natural language processing (NLP) are being pretrained on English text. When the parameters' number in the model gets more, the gap between English and other languages becomes harder to fill, and performance issues in other languages become problematic. To solve this issue, we present a new transfer learning approach that concentrates on knowledge transfer between two languages and makes a possible transfer to model size. After the transfer, we are looking to get a model of the same size as before. To make sure that model's size doesn't become a bottleneck in our approach, we train from scratch a model in our target language with a smaller size. After that, we use that model combined with the initial source model to construct token embeddings for the target model (which should be at the same size as source model) by contacting the vocabulary of both languages. The rest of the weights in the target model are same as in source model. This approach achieves the same outperforms standard language transfer method and gets 4 times faster convergence. [ABSTRACT FROM AUTHOR]

The Convolutional Sparse Coding (CSC) model has recently attracted a lot of attention in the signal and image processing communities. Since, in traditional sparse coding methods, a significant assumption is that all input samples are independent, so it is not well for most dependent works. In such cases, CSC models are a good choice. In this paper, we proposed a novel CSC-based classification model which combines the local block coordinate descent (LoBCoD) algorithm with the classification strategy. For this, in the training phase, the convolutional dictionary atoms (filters) of each class are learned by all training samples of the same class. In the test phase, the label of the query sample can be determined based on the reconstruction error of the filters related to every subject. Experimental results on five benchmark databases at the different number of training samples clearly demonstrate the superiority of our method to many state-of-the-art classification methods. Besides, we have shown that our method is less dependent on the number of training samples and therefore it can better work than other methods in small databases with fewer samples. For instance, increases of 26.27%, 18.32%, 11.35%, 13.5%, and 19.3% in recognition rates are observed for our method when compared to conventional SRC for five used databases at the least number of training samples per class. [ABSTRACT FROM AUTHOR]

The 'black box' characteristics of Convolutional Neural Networks (CNNs) present significant risks to their application scenarios, such as reliability, security, and division of responsibilities. Addressing the interpretability of CNN emerges as an urgent and critical issue in the field of machine learning. Recent research on CNN interpretability has either yielded unstable or inconsistent interpretations, or produced coarse-scale interpretable heatmaps, limiting their applicability in various scenarios. In this work, we propose a novel method of CNNs interpretation by incorporating a joint evaluation of multiple feature maps and employing multi-objective optimization (JE&MOO-CAM). Firstly, a method of joint evaluation for all feature maps is proposed to preserve the complete object instances and improve the overall activation values. Secondly, an interpretation method of CNNs under the MOO framework is proposed to avoid the instability and inconsistency of interpretation. Finally, the operators of selection, crossover, and mutation, along with the method of population initialization in NSGA-II, are redesigned to properly express the characteristics of CNNs. The experimental results, including both qualitative and quantitative assessments along with a sanity check conducted on three classic CNN models—VGG16, AlexNet, and ResNet50—demonstrate the superior performance of the proposed JE&MOO-CAM model. This model not only accurately pinpoints the instances within the image requiring explanation but also preserves the integrity of these instances to the greatest extent possible. These capabilities signify that JE&MOO-CAM surpasses six other leading state-of-the-art methods across four established evaluation criteria. [ABSTRACT FROM AUTHOR]

Abstract The feature map is a characteristic of high computational efficiency, but it is seldom used in path planning due to its lack of expression of environmental details. To solve this problem, a global path planning algorithm based on the feature map is proposed based on the directionality of line segment features. First, the robot searches the path along the direction of the target position but turns to search in the direction parallel to the obstacle, which it approaches until the line between the robot and the target position does not intersect with obstacles. Then it turns to the target position, keep searching the path. Meanwhile, the problems of the direction selection of turning point, corner point and obstacle circumvention in the searching process are analysed and corresponding solutions are put forth. Finally, a path optimisation algorithm with variable parameters is proposed, making the optimised path shorter and smoother. Simulation experiments demonstrate that the proposed algorithm is superior to A* algorithm in terms of computation time and path length, especially of the computation efficiency. The cover image is based on the Research Article A global path planning algorithm based on the feature map by Peng Liu et al., https://doi.org/10.1049/csy2.12040.

Unconstrained video face recognition is an extension of face recognition technology, and it is an indispensable part of intelligent security and criminal investigation systems. However, general face recognition technology cannot be directly applied to unconstrained video face recognition, because the video contains fewer frontal face image frames and a single image contains less face feature information. To address the above problems, this work proposes a Feature Map Aggregation Network (FMAN) to achieve unconstrained video face recognition by aggregating multiple face image frames. Specifically, an image group is used as the input of the feature extraction network to replace a single image to obtain a multi-channel feature map group. Then a quality perception module is proposed to obtain quality scores for feature maps and adaptively aggregate image features from image groups at the feature map level. Finally, extensive experiments are conducted on the challenging face recognition benchmarks YTF, IJB-A and COX to evaluate the proposed method, showing a significant increase in accuracy compared to the state-of-the-art. [ABSTRACT FROM AUTHOR]

Protein-protein interactions play an important role in life activities. The study of protein-protein interactions helps to better understand the mechanism of protein complex interaction, which is crucial for drug design, protein function annotation and three-dimensional structure prediction of protein complexes. In this paper, we study the tetramer protein complex interaction. The research has two parts: The first part is to predict the interaction between chains of the tetramer protein complex. In this part, we proposed a feature map to represent a sample generated by two chains of the tetramer protein complex, and constructed a Convolutional Neural Network (CNN) model to predict the interaction between chains of the tetramer protein complex. The AUC value of testing set is 0.6263, which indicates that our model can be used to predict the interaction between chains of the tetramer protein complex. The second part is to predict the tetramer protein complex interface residue pairs. In this part, we proposed a Support Vector Machine (SVM) ensemble method based on under-sampling and ensemble method to predict the tetramer protein complex interface residue pairs. In the top 10 predictions, when at least one protein-protein interaction interface is correctly predicted, the accuracy of our method is 82.14%. The result shows that our method is effective for the prediction of the tetramer protein complex interface residue pairs. [ABSTRACT FROM AUTHOR]

As an irreplaceable quantitative interpretation method, prestack seismic inversion enables the effective estimation of subsurface elastic parameters for reservoir prediction. However, for the model-driven prestack seismic inversion, the band-limited characteristics and noise interference of observed seismic data result in its high dependence on the initial models. This suggests that reasonable initial models act as a supplement to reliable variation trends in formation and can reduce the non-uniqueness of inversion results. In this article, we introduce a well-log interpolation method with a feature map-guided non-local means algorithm, which is for establishing high-fidelity initial models used for prestack seismic inversion. First, we briefly review the basic theory of general model-driven prestack seismic inversion. Then, we use dictionary learning to split the poststack seismic record into patches, and represent them with sparse vectors, instead of directly using seismic record. The advantage of dictionary learning is that it can adaptively extract useful signals from noisy observed data and provide fine structures by sparse reconstruction. Therefore, the proposed feature extraction method can improve the noise immunity and reliability of the well-log interpolation. More accurate initial models are pre-constructed efficiently by our feature extraction method, which improves the reliability of prestack seismic inversion results. Two kinds of observed seismic data are used, including the poststack seismic record for well-log interpolation and prestack seismic data used for inversion. Synthetic and field data tests both demonstrate the favorable performance of the proposed well-log interpolation method. In summary, a novel and convenient initial model building approach is provided, which contributes to seismic exploration and geologic modeling. [ABSTRACT FROM AUTHOR]

Objective: The diagnosis of prostate cancer has been a challenging task. Compared with traditional diagnosis methods, the radiofrequency (RF) signal is not only non-invasive but also rich in microscopic lesion information. This paper proposes a novel and accurate method for detecting prostate cancer based on the ultrasound RF signal. Method: Our approach is based on low-dimensional features in the frequency domain and high-throughput features in the spatial domain. The whole process could be divided into two parts: first, we calculate three feature maps from the ultrasound original RF signal, and 1,050 radiomics features are extracted from the three feature maps; second, we extracted 37 spectral features from the normalized frequency spectrum after Fourier transform. Results: We use LASSO regression as the method for feature selection; moreover, we use support vector machine (SVM) for classification 10-fold cross-validation for examining the classification performance of the SVM. An AUC (area under the receiver operating characteristic curve) of 0.84 was obtained on 71 subjects. Conclusions: Our method is feasible to detect prostate cancer based on the ultrasound RF signal with superior classification performance. [ABSTRACT FROM AUTHOR]