High-resolution Network Research Articles

Iris-LAHNet: a lightweight attention-guided high-resolution network for iris segmentation and localization

Iris-LAHNet: a lightweight attention-guided high-resolution network for iris segmentation and localization

Model for Identification of Electrical Appliance and Determination of Patterns Using High-Resolution Wireless Sensor NETWORK for the Efficient Home Energy Consumption Based on Deep Learning

The growing demand for electricity and the constant increase in electricity rates have intensified the interest of residential and non-residential energy consumers to reduce their energy consumption. The introduction of non-conventional renewable energies (photovoltaic and wind, in the residential case) demands new proposals to obtain a home energy management system (HEMS), which allows reducing the use of electrical energy. This article incorporates artificial intelligence techniques to demand response, allowing control, switching, turning on and off of appliances, modifying and reducing consumption, and achieving improvements in the quality of life in the home. In addition, an architecture based on a smart socket and an artificial intelligence model that recognizes the consumption of electrical appliances in high resolution (sampling every 10 s) is proposed. The system uses the Wi-Fi communication protocol, ensuring that the smart sockets wirelessly provide the data obtained to the public cloud. The use of Deep Learning allows us to obtain a central control model of the home, which, when interconnected to the smart electrical distribution networks of companies, could generate a positive impact on the environmental effects and CO2 reduction.

A rapid identification method for soft tissue markers of dentofacial deformities based on heatmap regression

ObjectiveThe purpose of this study was to construct a facial deformity dataset and a network model based on heatmap regression for the recognition of facial soft tissue landmarks to provide a basis for clinicians to perform cephalometric analysis of soft tissue.Materials and methodsA 34-point face marker detection model, the Back High-Resolution Network (BHR-Net), was constructed based on the heatmap regression algorithm, and a custom dataset of 1780 facial detection images for orthognathic surgery was collected. The mean normalized error (MNE) and 10% failure rate (FR10%) were used to evaluate the performance of BHR-Net, and a test set of 50 patients was used to verify the accuracy of the landmarks and their measurement indicators. The test results were subsequently validated in 30 patients.ResultsBoth the MNE and FR10% of BHR-Net were optimal compared with other models. In the test set (50 patients), the accuracy of the markers excluding the nose root was 86%, and the accuracy of the remaining markers reached 94%. In the model validation (30 patients), using the markers detected by BHR-Net, the diagnostic accuracy of doctors was 100% for Class II and III deformities, 100% for the oral angle plane, and 70% for maxillofacial asymmetric deformities.ConclusionsBHR-Net, a network model based on heatmap regression, can be used to effectively identify landmarks in maxillofacial multipose images, providing a reliable way for clinicians to perform cephalometric measurements of soft tissue objectively and quickly.

Spatially varying defocus map estimation from a single image based on spatial aliasing sampling method.

In current optical systems, defocus blur is inevitable due to the constrained depth of field. However, it is difficult to accurately identify the defocus amount at each pixel position as the point spread function changes spatially. In this paper, we introduce a histogram-invariant spatial aliasing sampling method for reconstructing all-in-focus images, which addresses the challenge of insufficient pixel-level annotated samples, and subsequently introduces a high-resolution network for estimating spatially varying defocus maps from a single image. The accuracy of the proposed method is evaluated on various synthetic and real data. The experimental results demonstrate that our proposed model outperforms state-of-the-art methods for defocus map estimation significantly.

Efficient super-resolution image reconstruction of electrical capacitance tomography for gas–solid fluidized bed measurement

Electrical capacitance tomography (ECT) has been widely used to investigate the flow dynamics of the gas–solid fluidized bed. Deep learning-based tomographic imaging of ECT is of great potential to reconstruct the particle concentration distribution with compatible image quality and reconstruction speed. However, it suffers from low model construction efficiency as numerous data samples need to be collected for training the deep neural network. Here a super-resolution (SR) imaging model based on a two-stage imaging network is proposed to improve the model construction efficiency. A low-resolution (LR) imaging network is used to reconstruct the LR image with the measured capacitance matrix while a subsequent high-resolution (HR) network calculates the HR image with the reconstructed LR image. The imaging performance, construction efficiency, and generalization ability of the trained SR imaging model are evaluated by reconstructing the samples in the test set and the unseen typical flow patterns. Considering the training set includes 100,000 data samples, the time consumption of numerically collecting the dataset for training the SR imaging model is 2.70 h, 2.2 % of that for a conventional deep learning-based imaging model of ECT, and the overall model construction time is only 4.44 h when the grid numbers of the LR and HR images are 246 and 8053, respectively. Experimental measurements are further carried out on a fluidized bed and the preliminary results show good prediction and generality of the trained model for reconstructing the particle concentration distribution.

Automatic Tracking of Hyoid Bone Displacement and Rotation Relative to Cervical Vertebrae in Videofluoroscopic Swallow Studies Using Deep Learning.

The hyoid bone displacement and rotation are critical kinematic events of the swallowing process in the assessment of videofluoroscopic swallow studies (VFSS). However, the quantitative analysis of such events requires frame-by-frame manual annotation, which is labor-intensive and time-consuming. Our work aims to develop a method of automatically tracking hyoid bone displacement and rotation in VFSS. We proposed a full high-resolution network, a deep learning architecture, to detect the anterior and posterior of the hyoid bone to identify its location and rotation. Meanwhile, the anterior-inferior corners of the C2 and C4 vertebrae were detected simultaneously to automatically establish a new coordinate system and eliminate the effect of posture change. The proposed model was developed by 59,468 VFSS frames collected from 1488 swallowing samples, and it achieved an average landmark localization error of 2.38 pixels (around 0.5% of the image with 448 × 448 pixels) and an average angle prediction error of 0.065 radians in predicting C2-C4 and hyoid bone angles. In addition, the displacement of the hyoid bone center was automatically tracked on a frame-by-frame analysis, achieving an average mean absolute error of 2.22 pixels and 2.78 pixels in the x-axis and y-axis, respectively. The results of this study support the effectiveness and accuracy of the proposed method in detecting hyoid bone displacement and rotation. Our study provided an automatic method of analyzing hyoid bone kinematics during VFSS, which could contribute to early diagnosis and effective disease management.

FSKT‐GE: Feature maps similarity knowledge transfer for low‐resolution gaze estimation

AbstractThe limited of texture details information in low‐resolution facial or eye images presents a challenge for gaze estimation. To address this, FSKT‐GE (feature maps similarity knowledge transfer for low‐resolution gaze estimation) is proposed, a gaze estimation framework consisting of both a high resolution (HR) network and low resolution (LR) network with the identical structure. Rather than mere feature imitation, this issue is addressed by assessing the cosine similarity of feature layers, emphasizing the distribution similarity between the HR and LR networks. This enables the LR network to acquire richer knowledge. This framework utilizes a combination loss function, incorporating cosine similarity measurement, soft loss based on probability distribution difference and gaze direction output, along with a hard loss from the LR network output layer. This approach on low‐resolution datasets derived from Gaze360 and RT‐Gene datasets is validated, demonstrating excellent performance in low‐resolution gaze estimation. Evaluations on low‐resolution images obtained through 2×, 4×, and 8× down‐sampling are conducted on two datasets. On the Gaze360 dataset, the lowest mean angular errors of 10.97°, 11.22°, and 13.61° were achieved, while on the RT‐Gene dataset, the lowest mean angular errors of 6.73°, 6.83°, and 7.75° were obtained.

Affordable event and monthly rain samplers: Improving isotopic datasets to understand meteorological processes.

Water-stable isotopes in rainfall are powerful tracers of atmospheric processes at different spatial and temporal scales. However, commercially available rain samplers for isotopic analysis are prohibitively expensive, especially for high spatial resolution networks and studies conducted in developing countries. A low-cost, simple, and robust sampler was designed for event and monthly rainfall samplings. Rainfall collectors were built based on existing designs provided in the literature and using easily accessible materials. Event samplers were filled with different volumes of reference water and left for 72 h in laboratory conditions to determine the minimum amount of rainfall to be collected to minimize isotopic fractionation, from both postsampling evaporation and equilibration. Samples were analyzed using dual-inlet isotope ratio mass spectrometry and cavity ring-down spectroscopy. For samples larger than 4% of the bottle's capacity, the evaporative enrichment due to Rayleigh distillation is negligible compared to the overall analytical uncertainty. Using a tube connecting the funnel to the water sample has proved to reduce postsampling evaporation by at least five times. To limit water self-diffusion, we recommend collecting the largest rainfall amount possible. Under these conditions, these collectors are suitable for rainfall sampling for isotopic analysis. This low-cost methodology will enable isotopic sampling of precipitation at high spatial resolutions and democratize the use of isotopes for hydrological studies in developing countries. All instructions for building and using these samplers are made openly accessible to the scientific community so they can be repeated and adapted to the needs of each project.

LION: Fast and High-Resolution Network Kernel Density Visualization

Network Kernel Density Visualization (NKDV) has often been used in a wide range of applications, e.g., criminology, transportation science, and urban planning. However, NKDV is computationally expensive, which cannot be scalable to large-scale datasets and high resolution sizes. Although a recent work, called aggregate distance augmentation (ADA), has been developed for improving the efficiency to generate NKDV, this method is still slow and does not take the resolution size into account for optimizing the efficiency. In this paper, we develop a new solution, called LION, which can reduce the worst-case time complexity for generating high-resolution NKDV, without increasing the space complexity. Experiment results on four large-scale location datasets verify that LION can achieve 2.86x to 35.36x speedup compared with the state-of-the-art ADA method.

Semantic segmentation of urban land classes using a multi-scale dataset

ABSTRACT The use of remote sensing imagery for land cover and land use classification has made significant advancements in recent years. However, it becomes particularly challenging to enhance the semantic representation of high-resolution networks while dealing with uneven land categories and merging multi-scale data without compromising the accuracy of semantic segmentation. To tackle this challenge, this paper presents a novel method for classifying high-resolution remote sensing images based on a deep neural network that performs semantic segmentation of urban construction lands into five categories: vegetation, water, buildings, roads, and bare soil. The network incorporates a U-shaped high-resolution neural network and the advanced high-resolution network (HRNet) framework. The parallel storage of feature maps with different resolutions enables the exchange of information between them. The data pre-processing module addresses the issue of data imbalance in the semantic segmentation of urban construction lands, resulting in an increase in Intersection over Union (IoU) values for different land types by 3.75%-12.01%. Additionally, a target context representation module is introduced to enhance the feature representation of pixels by calculating the relationship between pixels and multiple target regions. Moreover, a polarization attention mechanism is proposed to extract the characteristics of geographical objects in all directions and achieve a stronger semantic representation. This method provides a novel approach to accurately and effectively extract information on construction lands and advance the development of monitoring algorithms for urban construction lands. To validate the proposed U-HRNet-OCR+PSA network, a comparative analysis was conducted with six classical networks, including DeepLabv3+, PSPNet, U-Net, U-Net++, HRNet, and HRNet-OCR, as well as the relatively new ViT-adapter-L, Oneformer and InternImage-H. The experiments demonstrate that the U-HRNet-OCR+PSA network achieves higher accuracy compared to the aforementioned networks. Specifically, the corresponding IoU values for the buildings, roads, vegetation, bare soil, and water in the multi-scale dataset are 89.79%, 90.05%, 94.89%, 85.91%, and 88.36%, respectively.

Human Pose Estimation Based on Efficient and Lightweight High-Resolution Network (EL-HRNet).

As an important direction in computer vision, human pose estimation has received extensive attention in recent years. A High-Resolution Network (HRNet) can achieve effective estimation results as a classical human pose estimation method. However, the complex structure of the model is not conducive to deployment under limited computer resources. Therefore, an improved Efficient and Lightweight HRNet (EL-HRNet) model is proposed. In detail, point-wise and grouped convolutions were used to construct a lightweight residual module, replacing the original 3 × 3 module to reduce the parameters. To compensate for the information loss caused by the network's lightweight nature, the Convolutional Block Attention Module (CBAM) is introduced after the new lightweight residual module to construct the Lightweight Attention Basicblock (LA-Basicblock) module to achieve high-precision human pose estimation. To verify the effectiveness of the proposed EL-HRNet, experiments were carried out using the COCO2017 and MPII datasets. The experimental results show that the EL-HRNet model requires only 5 million parameters and 2.0 GFlops calculations and achieves an AP score of 67.1% on the COCO2017 validation set. In addition, PCKh@0.5mean is 87.7% on the MPII validation set, and EL-HRNet shows a good balance between model complexity and human pose estimation accuracy.

Precision in Motion: enhancing autonomous driving with advanced lane recognition using high resolution network

Precision in Motion: enhancing autonomous driving with advanced lane recognition using high resolution network

WideHRNet: An Efficient Model for Human Pose Estimation Using Wide Channels in Lightweight High-Resolution Network

WideHRNet: An Efficient Model for Human Pose Estimation Using Wide Channels in Lightweight High-Resolution Network

Self-Information Forgery Mining for Face Forgery Detection

Abstract In the face of rapid advances in face forgery technology, effective detection methods have become crucial to maintain the authenticity of digital media. Deep learning technology has provided new strategies for recognizing and preventing face forgery in recent years. In this study, a new face forgery detection technique is proposed by utilizing self-information theory, which improves the accuracy and robustness of detection by mining forgery traces, especially in diverse forgery scenarios. The study extracts face features through an improved high-resolution network HRNet and optimizes identity information extraction by combining facial reenactment techniques to detect forged faces efficiently. Experiments have been conducted on several mainstream forged face datasets, and the method presented in this paper can effectively improve the detection performance with an average accuracy of 74.75% on C40 recompressed images. Comparison experiments show that this research method’s frame-level and video-level detection accuracy on the Celeb-DF dataset are 0.9846 and 0.9985, respectively, which are higher than those of existing techniques. Cross-library tests validate the method’s generalization performance, and the AUC metric remains at 0.7305 even in low-quality video environments, which shows good resistance to environmental interference. This study proposes a self-information forgery mining technique that enhances forgery detection accuracy while demonstrating superior generalization ability.

Remote Sensing Monitoring of Cultivated Land Buildings Integration of High-Resolution Network and Dual-Branch Structure

Remote Sensing Monitoring of Cultivated Land Buildings Integration of High-Resolution Network and Dual-Branch Structure

One step accurate phase demodulation from a closed fringe pattern with the convolutional neural network HRUnet.

Retrieving a phase map from a single closed fringe pattern is a challenging task in optical interferometry. In this paper, a convolutional neural network (CNN), HRUnet, is proposed to demodulate phase from a closed fringe pattern. The HRUnet, derived from the Unet model, adopts a high resolution network (HRnet) module to extract high resolution feature maps of the data and employs residual blocks to erase the gradient vanishing in the network. With the trained network, the unwrapped phase map can be directly obtained by feeding a scaled fringe pattern. The high accuracy of the phase map obtained from HRUnet is demonstrated by demodulation of both simulated data and actual fringe patterns. Compared results between HRUnet and two other CNNS are also provided, and the results proved that the performance of HRUnet in accuracy is superior to the two other counterparts.

Dynamic context modeling based lightweight high-resolution network for dense prediction

Recent research shows that high-resolution networks can provide representative multi-scale features for vision tasks. However, high-resolution-based architectures are weak in capturing spatial long-range information, and they are computationally expensive. To alleviate these problems and encourage the engineering applications of dense prediction, based on vanilla high-resolution network, this paper presents a novel efficient architecture for dense prediction, namely Dynamic context modeling based lightweight High-Resolution Network (Dynamic-HRNet). In particular, the network consists of two key components: (i) dynamic split convolution, a more efficient and flexible convolution compared to the standard convolution, which could be conveniently embedded into other networks, and (ii) adaptive context modeling, which adaptively captures local and global contextual information to enrich the representation in parallel. Using the above two components, two lightweight blocks are designed that serve as the basic building units of Dynamic-HRNet, termed dynamic multi-scale context block and dynamic global context block, respectively. Extensive experiments demonstrate that the proposed Dynamic-HRNet, as a backbone, achieves significant superiority in popular benchmarks for human pose estimation (70.6% AP on COCO dataset and 87.6% PCKh on MPII dataset) and semantic segmentation (75.3% MIoU on Cityscapes dataset). Considering both efficiency and accuracy, it outperforms the most advanced lightweight architectures.

Automatic axial vertebral rotation estimation on radiographs for adolescent idiopathic scoliosis by deep learning

PurposeAdolescent idiopathic scoliosis (AIS) is a three-dimensional spine deformity governed by lateral curvature and axial vertebral rotation (AVR). Estimating AVR is important for treatment decisions and the prediction of AIS progression. However, manual AVR measurements have intra-observer and inter-observer errors, and therefore this paper proposes an automatic AVR measurement method to address the low precision issue of manual measurement. MethodWe develop an improved feature extraction module for vertebral landmark detection and pedicle segmentation. The improved coordinate convolution layer combined Polarized Self-Attention mechanism is applied in the feature extraction module to help the High-resolution Network to extract coordinate information. Based on vertebral landmark detection and pedicle segmentation, we propose an automatic AVR measurement algorithm to estimate the AVR. ResultsThe mean radial error (MRE) of vertebral landmark detection is 2.70 mm pixels, and the dice coefficient of pedicle segmentation is 72.45%. Compared with the original model, the MRE decreases by 2.13 mm, and the dice coefficient improves by 4.61%. For the AVR measurement results, the testing set contains 37 spine radiographs, including 481 vertebrae and 962 AVR measurements (481 vertebrae by two observers). The average measurement error is lower than 5°, which is within the standard of clinical measurement error. ConclusionThe results demonstrate that the proposed network performs well in vertebral landmark detection and pedicle segmentation, and the proposed AVR measurement is adopted for clinic diagnosis of AIS. SignificanceOur method achieves automatic AVR measurement, reducing the error introduced by manual measurement and improving the efficiency of orthopedists.

A High-Resolution Network with Strip Attention for Retinal Vessel Segmentation

Accurate segmentation of retinal vessels is an essential prerequisite for the subsequent analysis of fundus images. Recently, a number of methods based on deep learning have been proposed and shown to demonstrate promising segmentation performance, especially U-Net and its variants. However, tiny vessels and low-contrast vessels are hard to detect due to the issues of a loss of spatial details caused by consecutive down-sample operations and inadequate fusion of multi-level features caused by vanilla skip connections. To address these issues and enhance the segmentation precision of retinal vessels, we propose a novel high-resolution network with strip attention. Instead of the U-Net-shaped architecture, the proposed network follows an HRNet-shaped architecture as the basic network, learning high-resolution representations throughout the training process. In addition, a strip attention module including a horizontal attention mechanism and a vertical attention mechanism is designed to obtain long-range dependencies in the horizontal and vertical directions by calculating the similarity between each pixel and all pixels in the same row and the same column, respectively. For effective multi-layer feature fusion, we incorporate the strip attention module into the basic network to dynamically guide adjacent hierarchical features. Experimental results on the DRIVE and STARE datasets show that the proposed method can extract more tiny vessels and low-contrast vessels compared with existing mainstream methods, achieving accuracies of 96.16% and 97.08% and sensitivities of 82.68% and 89.36%, respectively. The proposed method has the potential to aid in the analysis of fundus images.

Edge Intelligence Empowered Vehicle Detection and Image Segmentation for Autonomous Vehicles

Edge intelligence (EI) migrates data and artificial intelligence (AI) to the “edge” of a network, enhancing the high-bandwidth and low-latency of wireless data transmission with the multiplier effect of 5G and AI, greatly improving the edges’ processing speed. Through integrating EI and computer vision technology, video surveillance systems in ITS can improve the processing capability of traffic information, which improves traffic efficiency and ensures traffic safety. Accordingly, first, we propose an edge intelligence-based improved-YOLOv4 vehicle detection algorithm, introducing an efficient channel attention (ECA) mechanism and a high-resolution network (HRNet) to enhance vehicle detection ability. Second, an edge intelligence-based improved DeepLabv3+ image segmentation algorithm is proposed, replacing the original backbone network with MobileNetv2 and using the softpool method, thus reducing the network size while improving the segmentation accuracy. Experimental results show that our proposed model has a higher average precision (AP) and can improve vehicle detection accuracy from 82.03% to 86.22%. The mean intersection over union (mIOU) of the image segmentation model improves from 73.32% to 75.63%.