Network Receptive Field Research Articles

Improving Differential-Neural Cryptanalysis

Our first objective is to enhance the capabilities of differential-neural distinguishers by applying more deep-learning techniques, focusing on handling more rounds and improving accuracy. Inspired by the Inception Block in GoogLeNet, we adopted a design that uses multiple parallel convolutional layers with varying kernel sizes before the residual block to capture multi-dimensional information. Additionally, we expanded the convolutional kernels in the residual blocks, enlarging the network's receptive field. In the case of Speck32/64, our efforts yield accuracy improvements in rounds 6, 7, and 8, enabling the successful training of a 9-round differential-neural distinguisher. As for Simon32/64, we developed a differential-neural distinguisher capable of effectively handling 12 rounds while achieving noteworthy accuracy enhancements in rounds 9, 10, and 11. Additionally, we utilized neutral bits to ensure the required data distribution for launching a successful key recovery attack when using multiple-ciphertext pairs as input for the neural network. Meanwhile, we redefined the formula for time complexity based on the differences in prediction speeds of the distinguisher between a single-core CPU and a GPU. Combining these various advancements allows us to considerably reduce the time and data complexity of key recovery attacks on 13-round Speck32/64. Furthermore, we used knowledge distillation techniques to reduce the model size, accelerating the distinguisher's prediction speed and reducing the time complexity. In particular, we achieved a successful 14-round key recovery attack by exhaustively guessing a 1-round subkey. For Simon32/64, we accomplished a 17-round key recovery attack for the first time and reduced the time complexity of the 16-round key recovery attack.

Embedded U-shaped network with cross-hierarchical feature adaptation fusion for remote sensing image haze removal

ABSTRACT Recently, U-Net architecture has been extensively explored for remote sensing (RS) image haze removal, demonstrating remarkable performance. However, most existing RS image haze removal methods based on the U-Net fail to fully utilize feature information extracted in the encoding layers during decoding, resulting in unsatisfactory haze removal results. Moreover, most haze removal algorithms neglect the interaction of cross-level feature information, which is particularly important for scene recovery and context information constraints. To address these issues, this paper presents an Embedded U-shaped Network with Cross-Hierarchical Feature Adaptation Fusion Network (EUCHA). Specifically, an Embedded U-shaped Network Framework (EUNF) is proposed to fuse the features extracted from different scale encoding layers as supplementary information into the corresponding decoding layers by embedding multiple parallel U-shaped subnetworks, which alleviates feature information dilution during the decoding process and enhances the network receptive field. Furthermore, we propose a Cross-Hierarchical Feature Adaptive Fusion (CHAF) module, which effectively and adaptively fuses multiple adjacent hierarchical features extracted by channel attention and pixel attention through learnable factors to enhance the expressive ability of features. Experiments were conducted on both real and synthetic datasets and compared with the nine most advanced algorithms. The results showed that EUCHA algorithm performs better in removing artefacts in both RS images and natural images. The source code is available at https://github.com/GongHongY/EUCHA.

BMSeNet: Multiscale Context Pyramid Pooling and Spatial Detail Enhancement Network for Real-Time Semantic Segmentation.

Most real-time semantic segmentation networks use shallow architectures to achieve fast inference speeds. This approach, however, limits a network's receptive field. Concurrently, feature information extraction is restricted to a single scale, which reduces the network's ability to generalize and maintain robustness. Furthermore, loss of image spatial details negatively impacts segmentation accuracy. To address these limitations, this paper proposes a Multiscale Context Pyramid Pooling and Spatial Detail Enhancement Network (BMSeNet). First, to address the limitation of singular semantic feature scales, a Multiscale Context Pyramid Pooling Module (MSCPPM) is introduced. By leveraging various pooling operations, this module efficiently enlarges the receptive field and better aggregates multiscale contextual information. Moreover, a Spatial Detail Enhancement Module (SDEM) is designed, to effectively compensate for lost spatial detail information and significantly enhance the perception of spatial details. Finally, a Bilateral Attention Fusion Module (BAFM) is proposed. This module leverages pixel positional correlations to guide the network in assigning appropriate weights to the features extracted from the two branches, effectively merging the feature information of both branches. Extensive experiments were conducted on the Cityscapes and CamVid datasets. Experimental results show that the proposed BMSeNet achieves a good balance between inference speed and segmentation accuracy, outperforming some state-of-the-art real-time semantic segmentation methods.

Diagnosis of Sliding Bearing Lubrication State Based on SC-ResNet

In response to the issue of poor diagnostic performance of models due to the scarcity of samples for lubrication failure states in sliding bearings in practical engineering applications, this thesis proposes a fault diagnosis method for the lubrication state of sliding bearings based on a Self-Calibrated Residual Network (SC-ResNet). By leveraging the self-calibrated convolution, which significantly enhances the network's receptive field and feature extraction capabilities without increasing the number of parameters and complexity, a self-calibrated residual block is designed to construct the SC-ResNet model. This model is capable of diagnosing the lubrication state of bearings using a small number of samples inputted into a pre-trained model. Experimental results indicate that this method performs exceptionally well under conditions with a limited number of samples, achieving higher Recall and F1-score values compared to other methods.

Multi-attention fusion transformer for single-image super-resolution

Recently, Transformer-based methods have gained prominence in image super-resolution (SR) tasks, addressing the challenge of long-range dependence through the incorporation of cross-layer connectivity and local attention mechanisms. However, the analysis of these networks using local attribution maps has revealed significant limitations in leveraging the spatial extent of input information. To unlock the inherent potential of Transformer in image SR, we propose the Multi-Attention Fusion Transformer (MAFT), a novel model designed to integrate multiple attention mechanisms with the objective of expanding the number and range of pixels activated during image reconstruction. This integration enhances the effective utilization of input information space. At the core of our model lies the Multi-attention Adaptive Integration Groups, which facilitate the transition from dense local attention to sparse global attention through the introduction of Local Attention Aggregation and Global Attention Aggregation blocks with alternating connections, effectively broadening the network's receptive field. The effectiveness of our proposed algorithm has been validated through comprehensive quantitative and qualitative evaluation experiments conducted on benchmark datasets. Compared to state-of-the-art methods (e.g. HAT), the proposed MAFT achieves 0.09 dB gains on Urban100 dataset for × 4 SR task while containing 32.55% and 38.01% fewer parameters and FLOPs, respectively.

A new segmentation algorithm for peripapillary atrophy and optic disk from ultra-widefield Photographs

Background and objectiveThe prevalence of myopia and high myopia is increasing globally, underscoring the growing importance of diagnosing high myopia-related pathologies. While existing image segmentation models, such as U-Net, UNet++, ResU-Net, and TransUNet, have achieved significant success in medical image segmentation, they still face challenges when dealing with ultra-widefield (UWF) fundus images. This study introduces a novel automatic segmentation algorithm for the optic disc and peripapillary atrophy (PPA) based on UWF fundus images, aimed at assisting ophthalmologists in more accurately diagnosing high myopia-related diseases. MethodsIn this study, we developed a segmentation model leveraging a Transformer-based network structure, complemented by atrous convolution and selective boundary aggregation modules, to elevate the accuracy of segmenting the optic disc and PPA in UWF photography. The atrous convolution module adeptly manages multi-scale features, catering to the variances in target sizes and expanding the deep network's receptive field. Concurrently, the incorporation of the selective boundary aggregation module in the skip connections of the model significantly improves the differentiation of boundary information between segmentation targets. Moreover, the comparison of our proposed algorithm with classical segmentation models like U-Net, UNet++, ResU-Net, and TransUNet highlights its considerable advantages in processing UWF photographs. ResultsThe experimental results show that, compared to the other four models, our algorithm demonstrates substantial improvements in segmenting the optic disc and PPA in UWF photographs. In PPA segmentation, our algorithm improves by 0.8% in Dice, 1.8% in sensitivity, and 1.3% in intersection over union (IOU). In optic disc segmentation, our algorithm improves by 0.3% in Dice, 0.6% in precision, and 0.4% in IOU. ConclusionOur proposed method improves the segmentation accuracy of PPA and optic disks based on UWF photographs, which is valuable for diagnosing high myopia-related diseases in ophthalmology clinics.

HIMS-Net: Horizontal-vertical interaction and multiple side-outputs network for cyst segmentation in jaw images.

Jaw cysts are mainly caused by abnormal tooth development, chronic oral inflammation, or jaw damage, which may lead to facial swelling, deformity, tooth loss, and other symptoms. Due to the diversity and complexity of cyst images, deep-learning algorithms still face many difficulties and challenges. In response to these problems, we present a horizontal-vertical interaction and multiple side-outputs network for cyst segmentation in jaw images. First, the horizontal-vertical interaction mechanism facilitates complex communication paths in the vertical and horizontal dimensions, and it has the ability to capture a wide range of context dependencies. Second, the feature-fused unit is introduced to adjust the network's receptive field, which enhances the ability of acquiring multi-scale context information. Third, the multiple side-outputs strategy intelligently combines feature maps to generate more accurate and detailed change maps. Finally, experiments were carried out on the self-established jaw cyst dataset and compared with different specialist physicians to evaluate its clinical usability. The research results indicate that the Matthews correlation coefficient (Mcc), Dice, and Jaccard of HIMS-Net were 93.61, 93.66 and 88.10% respectively, which may contribute to rapid and accurate diagnosis in clinical practice.

FSE-Net: Rethinking the up-sampling operation in encoder-decoder structure for retinal vessel segmentation

Automatic retinal vessel segmentation plays a crucial role in the diagnosis and assessment of various ophthalmologic diseases. Currently, the primary retinal vessel segmentation algorithms are based on the encoder-decoder structure. However, these U-Net analogs suffer from the loss of both spatial and semantic information, caused by continuous up-sampling operations in the decoder structure. In this paper, we rethink the above problem and build a novel deep neural network for retinal vessel segmentation, called FSE-Net. Specifically, to address the issue of feature information loss and enhance the performance of retinal vessel segmentation, we eliminate the decoder structure. In particular, we introduced a multi-head feature fusion block (MFF) as a substitute for the continuous up-sampling operation. Additionally, the encoder stage of FSE-Net incorporates a residual feature separable block (RFSB) to further refine and distill features, thereby enhancing the capability of feature extraction. Subsequently, we employ a residual atrous spatial feature aggregate module (RASF) to expand the network's receptive field by incorporating multi-scale feature information. We conducted experiments on five widely recognized databases for retinal vessel segmentation, namely DRIVE, CHASEDB1, STARE, IOSTAR, and LES-AV. The results demonstrate that our proposed FSE-Net outperforms state-of-the-art approaches in terms of segmentation performance. Moreover, we demonstrate the feasibility of achieving superior segmentation performance without employing the traditional U-Net analog network structure.

Influence of Neural Network Receptive Field on Monocular Depth and Ego-Motion Estimation

We present an analysis of a self-supervised learning approach for monocular depth and ego-motion estimation. This is an important problem for computer vision systems of robots, autonomous vehicles and other intelligent agents, equipped only with monocular camera sensor. We have explored a number of neural network architectures that perform single-frame depth and multi-frame camera pose predictions to minimize photometric error between consecutive frames on a sequence of camera images. Unlike other existing works, our proposed approach called ERF-SfMLearner examines the influence of the deep neural network receptive field on the performance of depth and ego-motion estimation. To do this, we study the modification of network layers with two convolution operators with extended receptive field: dilated and deformable convolutions. We demonstrate on the KITTI dataset that increasing the receptive field leads to better metrics and lower errors both in terms of depth and ego-motion estimation. Code is publicly available at github.com/linukc/ERF-SfMLearner.

Real-time Monitoring System for Driver Phone Usage Based on Improved YOLOv5s

In response to the impact of driver's violations, such as using a mobile phone, on vehicle safety during the driving process, we propose an improved real-time monitoring algorithm based on YOLOv5s with lightweight optimization. Firstly, we replace the C3 module (CSP Bottleneck with 3 convolutions) in the backbone network of YOLOv5s with a lightweight Ghost Module to reduce the model's parameter count, enhance detection speed, and maintain inference accuracy unaffected, thus meeting the requirements of real-time monitoring. Secondly, we introduce the RepConv (Receptive Field Block) module into the Feature Extraction Network (PANet) structure to increase the neural network's receptive field for input images and further reduce the model's computational load. Experimental results show that the improved network achieves an mAP@0.5 of 95.7%, a detection speed of 140 FPS, and a model size reduction to 10.6MB, meeting the demand for real-time and reliable detection on embedded devices.

Single-image super-resolution based on an improved asymmetric Laplacian pyramid structure

Large-scale factor image super-resolution whose scale factor is greater than 4 is significant in real-world applications for single image super-resolution. The current image super-resolution techniques for large-scale factor, however, frequently upsample low-resolution images in a single pass, leading to edge artifacts in the reconstructed images. In this article, we provide an improved asymmetric Laplacian pyramid network to further realize large-scale factor image super-resolution and fully utilize various size features. A distinct architecture is applied at each layer of the pyramid for improving feature extraction. We extend the first layer of the pyramid with a lightweight transformer design, which enables the model to efficiently collect the contextual information in the sequence by utilizing the multi-headed attention mechanism. Additionally, we combine improved dense skip connections with recursive operations to form a deep dense recursive convolutional neural network that fuses low-level and high-level features while broadening the network's receptive field. The quantitative and qualitative analysis on benchmark dataset's demonstrate that our method provides superior performance in both PSNR and SSIM and is more in line with human visual perception.

Semantic Segmentation of Urban Airborne LiDAR Point Clouds Based on Fusion Attention Mechanism and Multi-Scale Features

Semantic segmentation of point clouds provided by airborne LiDAR survey in urban scenes is a great challenge. This is due to the fact that point clouds at boundaries of different types of objects are easy to be mixed and have geometric spatial similarity. In addition, the 3D descriptions of the same type of objects have different scales. To address above problems, a fusion attention convolutional network (SMAnet) was proposed in this study. The fusion attention module includes a self-attention module (SAM) and multi-head attention module (MAM). The SAM can capture feature information according to correlation of adjacent point cloud and it can distinguish the mixed point clouds with similar geometric features effectively. The MAM strengthens connections among point clouds according to different subspace features, which is beneficial for distinguishing point clouds at different scales. In feature extraction, lightweight multi-scale feature extraction layers are used to effectively utilize local information of different neighbor fields. Additionally, in order to solve the feature externalization problem and expand the network receptive field, the SoftMax-stochastic pooling (SSP) algorithm is proposed to extract global features. The ISPRS 3D Semantic Labeling Contest dataset was chosen in this study for point cloud segmentation experimentation. Results showed that the overall accuracy and average F1-score of SMAnet reach 85.7% and 75.1%, respectively. It is therefore superior to common algorithms at present. The proposed model also achieved good results on the GML(B) dataset, which proves that the model has good generalization ability.

Skin lesion recognition via global-local attention and dual-branch input network

Owing to the deficiency of training data, the variety of skin lesion shapes for different patients, inter-class similarity, and intra-class variation, the recognition accuracy and speed in skin lesion recognition based on deep learning remain challenging. To solve the above issues, we propose a Dual-Branch and Global-Local Attention network based on ResNet50 (DGLA-ResNet50) to reduce model parameters and improve identification accuracy. For inter-class and intra-class problems, a global-local attention mechanism is designed to obtain the dependency information of query points in horizontal and vertical directions in turn for obtaining the global information indirectly. And we choose to obtain the local information through multiple convolutions simultaneously. Aiming at insufficient samples, the attention mechanism is lightened from the perspectives of reducing query points and the points associated with them, so as to reduce parameter redundancy. In addition, we present a dual-branch input network to deal with the problem of image variety, in which two branches are used to extract image features with different resolutions for fusion, so as to expand the network receptive field. We evaluated DGLA-ResNet50 on ISIC2018 and ISIC2019. The experimental results on ISIC2018 demonstrate that DGLA-ResNet50 has a recognition accuracy of 90.71%, while the parameters of the model are only 104.2M and the FLOPs value is 15.6G. Compared with the common models, our model obtains significantly better performance. The results indicate that DGLA-ResNet50 can improve the accuracy well while ensuring the lightweight of the model, and can prospectively assist doctors in the rapid diagnosis of skin lesions.

CS-MRI Reconstruction Using an Improved GAN with Dilated Residual Networks and Channel Attention Mechanism.

Compressed sensing (CS) MRI has shown great potential in enhancing time efficiency. Deep learning techniques, specifically generative adversarial networks (GANs), have emerged as potent tools for speedy CS-MRI reconstruction. Yet, as the complexity of deep learning reconstruction models increases, this can lead to prolonged reconstruction time and challenges in achieving convergence. In this study, we present a novel GAN-based model that delivers superior performance without the model complexity escalating. Our generator module, built on the U-net architecture, incorporates dilated residual (DR) networks, thus expanding the network's receptive field without increasing parameters or computational load. At every step of the downsampling path, this revamped generator module includes a DR network, with the dilation rates adjusted according to the depth of the network layer. Moreover, we have introduced a channel attention mechanism (CAM) to distinguish between channels and reduce background noise, thereby focusing on key information. This mechanism adeptly combines global maximum and average pooling approaches to refine channel attention. We conducted comprehensive experiments with the designed model using public domain MRI datasets of the human brain. Ablation studies affirmed the efficacy of the modified modules within the network. Incorporating DR networks and CAM elevated the peak signal-to-noise ratios (PSNR) of the reconstructed images by about 1.2 and 0.8 dB, respectively, on average, even at 10× CS acceleration. Compared to other relevant models, our proposed model exhibits exceptional performance, achieving not only excellent stability but also outperforming most of the compared networks in terms of PSNR and SSIM. When compared with U-net, DR-CAM-GAN's average gains in SSIM and PSNR were 14% and 15%, respectively. Its MSE was reduced by a factor that ranged from two to seven. The model presents a promising pathway for enhancing the efficiency and quality of CS-MRI reconstruction.

A fault diagnosis method based on dilated convolution and attention for rolling bearing under multiple working conditions and noisy environments

As essential equipment in rotating machinery, the fault diagnosis technology of rolling bearings has achieved great success. However, it still suffers from limitations in terms of generalization and noise resistance performance when operating under complex conditions. To accurately identify the fault types of rolling bearings under different loads and nosy environments, a novel intelligent fault diagnosis method is proposed. Firstly, the utilization of dilated convolution expands the network's receptive field, thereby effectively enhancing the scope of fault extraction. Then, by incorporating the Efficient Channel Attention (ECA) in different convolutional layers, the extracted features are adaptively recognized, highlighting important representation information and improving fault diagnosis performance. Finally, the proposed network is utilized for rolling bearing fault diagnosis under diverse operating and noise conditions, and its efficacy is evaluated on various datasets. The experimental results demonstrate that the proposed method exhibits good generalization performance and strong robustness, compared with other methods.

Learning a Deep Attention Dilated Residual Convolutional Neural Network for Landslide Susceptibility Mapping in Hanzhong City, Shaanxi Province, China

The analysis and evaluation of landslide susceptibility are of great significance in preventing and managing geological hazards. Aiming at the problems of insufficient information caused by the limited number of landslide datasets, complex information of landslide evaluation factors, and low prediction accuracy of landslide susceptibility, a landslide susceptibility evaluation method based on the deep attention dilated residual convolutional neural network (DADRCNN) is proposed. First, the dilated convolution unit (DCU) is used to increase the network receptive field, aggregate multi-scale information, and enhance the model ability to capture the characteristics of landslide evaluation factors. Second, the deep residual module (DRM) is used to solve the issue of gradient disappearance and better extract data features by overlaying the residual function mapping layer and increasing the network depth. Finally, the channel attention residual module (CARM) is introduced to learn the varying importance of different landslide evaluation factors, and assign different weights to improve the susceptibility prediction accuracy. The experimental results show that the DADRCNN method can extract features around the sample points, expand the receptive field, and deeply mine the information. It mitigates the lack of sample information in training, focuses on important feature information, and significantly improves the prediction accuracy.

An improved image processing algorithm for visual characteristics in graphic design.

Drawing the clothing plan is an essential part of the clothing industry. However, the irregular shape of clothing, strong deformability and sensitivity to light make the fast and accurate realization of clothing image retrieval a very challenging problem. The successful application of the Transformer in image recognition shows the application potential of the Transformer in the image field. This article proposes an efficient and improved clothing plan based on ResNet-50. Firstly, in the feature extraction section, the ResNet-50 network structure embedded in the Transformer module is used to improve the network's receptive field range and feature extraction ability. Secondly, dense jump connections are added to the ResNet-50 upsampling process, making full use of feature extraction information at each stage, further improving the quality of the generated image. The network consists of three steps: the sketch stage, which aims to predict the color distribution of clothing and obtain watercolor images without gradients and shadows. The second is the thinning stage, which refines the watercolor image into a clothing image with light and shadow effect; The third is the optimization stage, which combines the outputs of the first two stages to optimize the generation quality further. The experimental results show that the improved network's IS and first input delay (FID) scores are 4.592 and 1.506, respectively. High-quality clothing images can be generated only by inputting line drawings and a few color points. Compared with the existing methods, the image generated by this network has excellent advantages in realism and accuracy. This method can combine various feature information of images, improve retrieval accuracy, has strong robustness and practicability, and can provide a reference for the daily work of fashion designers.

Remote Sensing Image Change Detection Network Based on Twin High-Resolution Representation

With the increase of spatial resolution of remote sensing images, features of feature imaging become more and more complex, and the change detection methods based on techniques such as texture representation and local semantics are difficult to meet the demand. Most change detection methods usually focus on extracting semantic features and ignore the importance of high-resolution shallow information and fine-grained features, which often lead to uncertainty in edge detection and small target detection. For single-input networks when two temporal images are connected, the shallow layer of the network cannot provide the information of the individual original image to the deep layer features to help reconstruct the image, and therefore, the change detection results may be missing in detail and feature compactness. For this purpose, a twins context aggregation network (TCANet) is proposed to perform change detection on remote sensing images. In order to reduce the loss of spatial accuracy of remote sensing images and maintain high-resolution representation, we introduce HRNet as our backbone network to initially extract the features of interest. Our proposed context aggregation module (CAM) can amplify the convolutional neural network receptive field to obtain more detailed contextual information without significantly increasing the computational effort. The side output embedding module (SOEM) is proposed to improve the accuracy of small volume target change detection as well as to shorten the training process and speed up the detection while ensuring the performance. The method has experimented on the publicly available CDD dataset, the SYSU-CD dataset, and a challenging DSIFN dataset. With significant improvements in precision, recall, F1 score, and overall accuracy, the method outperforms the five methods mentioned in the literature.

Multi‐task multi‐scale attention learning‐based facial age estimation

Face-based age estimation strongly depends on deep residual networks (ResNets), used as the backbone in the relevant research. However, ResNet-based methods ignore the importance of some large-scale facial information and other facial age attributes. Inspired by the attention mechanism, a multi-task learning framework for face-based age estimation called multi-task multi-scale attention is proposed. First, the authors embed the alternative strategy structure of dilated convolution into ResNet34 to construct a multi-scale attention module (MSA) to improve the network's receptive field, which extracts local age-sensitive information while obtaining multi-scale features. The MSA can have a larger receptive field to extract both large-scale and local detailed feature information. Second, multi-task learning network structures are built to predict gender and race, which can share rigid network parameters to improve age estimation and improve the accuracy of age estimation by other age-related parameters. Finally, the Kullback-Leibler divergence loss is adopted between a Dirac delta label and a Gaussian prediction to guide the training. The numerical tests on the MORPH Album II and Adience datasets prove the superiority of the proposed method over other state-of-the-art ones.

Deep Adaptive Sampling and Reconstruction Using Analytic Distributions

We propose an adaptive sampling and reconstruction method for offline Monte Carlo rendering. Our method produces sampling maps constrained by a user-defined budget that minimize the expected future denoising error. Compared to other state-of-the-art methods, which produce the necessary training data on the fly by composing pre-rendered images, our method samples from analytic noise distributions instead. These distributions are compact and closely approximate the pixel value distributions stemming from Monte Carlo rendering. Our method can efficiently sample training data by leveraging only a few per-pixel statistics of the target distribution, which provides several benefits over the current state of the art. Most notably, our analytic distributions' modeling accuracy and sampling efficiency increase with sample count, essential for high-quality offline rendering. Although our distributions are approximate, our method supports joint end-to-end training of the sampling and denoising networks. Finally, we propose the addition of a global summary module to our architecture that accumulates valuable information from image regions outside of the network's receptive field. This information discourages sub-optimal decisions based on local information. Our evaluation against other state-of-the-art neural sampling methods demonstrates denoising quality and data efficiency improvements.