Traditional U-Net Research Articles

G2D-Net: Efficient Dehazing with Second-Order Gated Units

Image dehazing aims to reconstruct potentially clear images from corresponding images corrupted by haze. With the rapid development of deep learning-related technologies, dehazing methods based on deep convolutional neural networks have gradually become mainstream. We note that existing dehazing methods often accompany an increase in computational overhead while improving the performance of dehazing. We propose a novel lightweight dehazing neural network to balance performance and efficiency: the g2D-Net. The g2D-Net borrows the design ideas of input-adaptive and long-range information interaction from Vision Transformers and introduces two kinds of convolutional blocks, i.e., the g2D Block and the FFT-g2D Block. Specifically, the g2D Block is a residual block with second-order gated units, which inherit the input-adaptive property of a gated unit and can realize the second-order interaction of spatial information. The FFT-g2D Block is a variant of the g2D Block, which efficiently extracts the global features of the feature maps through fast Fourier convolution and fuses them with local features. In addition, we employ the SK Fusion layer to improve the cascade fusion layer in a traditional U-Net, thus introducing the channel attention mechanism and dynamically fusing information from different paths. We conducted comparative experiments on five benchmark datasets, and the results demonstrate that the g2D-Net achieves impressive dehazing performance with relatively low complexity.

Breast Cancer Tumour Masks Segmentation Using Modified U-Net with Channel Attention

Abstract: A modified U-Net for biomedical image segmentation with channel attention is developed for breast cancer image segmentation. A unique step in biomedical image analysis is segmentation because it determines the accuracy of image analysis algorithms. Algorithms for accurate and early breast cancer detection in women using mammograms depend highly on the accuracy of the segmentation stage. In this thesis, in order to improve on the segmentation ability of the traditional UNet architecture, it was modified by varying the fixed kernel sizes that it was originally designed for. In this model, progressively increasing kernel sizes of 3x3, 5x5, 7x7, 9x9 and 11x11 were used, which is different from the architecture of UNet that depended on fixed 3x3 kernel sizes for all convolutional layers. The first Convolutional block of the proposed model contains two Conv2D layers with 3x3 kernel sizes, ReLU (Rectified Linear Unit) activation function, and 16 channel dimensions. A maxpooling2D layer with pool size of (2x2) was placed to downsample the image size before feeding to the next convolutional block which used the same activation function, with 32 channel dimension and 5x5 kernel size. The next convolutional block used 64 channel dimensions with 7x7 kernel size, then a 9x9 kernel size Conv2D block with 128 channel dimensions was placed before the Conv2D with upsampling. The squeeze and excitation block was placed after this convolutional layer and in-between the two convolutional blocks with 11x11 kernel sizes, to reallocate the weight of the essential features to ensure that more discriminate features are learnt from the breast cancer images to improve the overall representational strength of the network through the performance of dynamic feature recalibration of the image channels. This significantly improved the performance of the segmentation by explicitly modeling the interdependencies between channels in the convolutional layers. The increasing kernel sizes enables the model to learn more discriminate features from the images. Jaccard similarity index, F1 score, Recall, Precision, F2 score, and accuracy results were used to compare the result of the proposed model with other models such as U-Net model and Baseline model. The results shows a better performance of 0.88 for Jaccard similarity index, 0.94 for F1 score, 0.93 for recall, 0.92 for precision, 0.92 for F2 score, and an accuracy score of 0.98%. The model is good for biomedical image segmentation.

Imaging segmentation mechanism for rectal tumors using improved U-Net

ObjectiveIn radiation therapy, cancerous region segmentation in magnetic resonance images (MRI) is a critical step. For rectal cancer, the automatic segmentation of rectal tumors from an MRI is a great challenge. There are two main shortcomings in existing deep learning-based methods that lead to incorrect segmentation: 1) there are many organs surrounding the rectum, and the shape of some organs is similar to that of rectal tumors; 2) high-level features extracted by conventional neural networks often do not contain enough high-resolution information. Therefore, an improved U-Net segmentation network based on attention mechanisms is proposed to replace the traditional U-Net network.MethodsThe overall framework of the proposed method is based on traditional U-Net. A ResNeSt module was added to extract the overall features, and a shape module was added after the encoder layer. We then combined the outputs of the shape module and the decoder to obtain the results. Moreover, the model used different types of attention mechanisms, so that the network learned information to improve segmentation accuracy.ResultsWe validated the effectiveness of the proposed method using 3773 2D MRI datasets from 304 patients. The results showed that the proposed method achieved 0.987, 0.946, 0.897, and 0.899 for Dice, MPA, MioU, and FWIoU, respectively; these values are significantly better than those of other existing methods.ConclusionDue to time savings, the proposed method can help radiologists segment rectal tumors effectively and enable them to focus on patients whose cancerous regions are difficult for the network to segment.SignificanceThe proposed method can help doctors segment rectal tumors, thereby ensuring good diagnostic quality and accuracy.

Fast Lung Image Segmentation Using Lightweight VAEL-Unet

INTRODUCTION: A lightweght lung image segmentation model was explored. It was with fast speed and low resouces consumed while the accuracy was comparable to those SOAT models. OBJECTIVES: To improve the segmentation accuracy and computational efficiency of the model in extracting lung regions from chest X-ray images, a lightweight segmentation model enhanced with a visual attention mechanism called VAEL-Unet, was proposed. METHODS: Firstly, the bneck module from the MobileNetV3 network was employed to replace the convolutional and pooling operations at different positions in the U-Net encoder, enabling the model to extract deeper-level features while reducing complexity and parameters. Secondly, an attention module was introduced during feature fusion, where the processed feature maps were sequentially fused with the corresponding positions in the decoder to obtain the segmented image. RESULTS: On ChestXray, the accuracy of VAEL-Unet improves from 97.37% in the traditional U-Net network to 97.69%, while the F1-score increases by 0.67%, 0.77%, 0.61%, and 1.03% compared to U-Net, SegNet, ResUnet and DeepLabV3+ networks. respectively. On LUNA dataset. the F1-score demonstrates improvements of 0.51%, 0.48%, 0.22% and 0.46%, respectively, while the accuracy has increased from 97.78% in the traditional U-Net model to 98.08% in the VAEL-Unet model. The training time of the VAEL-Unet is much less compared to other models. The number of parameters of VAEL-Unet is only 1.1M, significantly less than 32M of U-Net, 29M of SegNet, 48M of Res-Unet, 5.8M of DeeplabV3+ and 41M of DeepLabV3Plus_ResNet50. CONCLUSION: These results indicate that VAEL-Unet’s segmentation performance is slightly better than other referenced models while its training time and parameters are much less.

MSPAN: Multi‐scale pyramid attention network for efficient skin cancer lesion segmentation

AbstractSkin cancer is common and deadly, needs to be detected and treated properly. Deep learning algorithms like UNet have shown potential results in medical imaging. Such approaches still struggle to capture fine‐grained details and scale differences in skin lesions‐based occlusions' appearance, size etc. This research proposes a redesign UNet, the Multi‐Scale Pyramid Attention Network (MSPAN), to improve skin cancer lesion segmentation. The input data is processed at numerous scales with varied receptive fields. This enhances the network's ability to identify lesion locations by capturing local and global context. Attention approaches also help the network to suppress noise by focusing on informative features. We have evaluated MSPAN model on the publicly available ISIC2018 benchmark dataset for skin lesion segmentation. The method surpasses traditional UNet and other current methods in accuracy and effectiveness. The model also has a post‐processing to estimate lesion area for fast inference, making it suitable for extensive screening. Redesigned UNet with the Multi‐Scale Pyramid Attention Network improves skin cancer lesion segmentation. The model's ability to collect fine‐grained information and handle occlusions allows for more accurate skin cancer diagnosis and treatment. The MSPAN design can improve computer‐aided diagnosis systems and help dermatologists make precise clinical decisions.

Multimodal Biomedical Image Segmentation using Multi-Dimensional U-Convolutional Neural Network

Deep learning recently achieved advancement in the segmentation of medical images. In this regard, U-Net is the most predominant deep neural network, and its architecture is the most prevalent in the medical imaging society. Experiments conducted on difficult datasets directed us to the conclusion that the traditional U-Net framework appears to be deficient in certain respects, despite its overall excellence in segmenting multimodal medical images. Therefore, we propose several modifications to the existing cutting-edge U-Net model. The technical approach involves applying a Multi-Dimensional U-Convolutional Neural Network to achieve accurate segmentation of multimodal biomedical images, enhancing precision and comprehensiveness in identifying and analyzing structures across diverse imaging modalities. As a result of the enhancements, we propose a novel framework called Multi-Dimensional U-Convolutional Neural Network (MDU-CNN) as a potential successor to the U-Net framework. On a large set of multimodal medical images, we compared our proposed framework, MDU-CNN, to the classical U-Net. There have been small changes in the case of perfect images, and a huge improvement is obtained in the case of difficult images. We tested our model on five distinct datasets, each of which presented unique challenges, and found that it has obtained a better performance of 1.32%, 5.19%, 4.50%, 10.23% and 0.87%, respectively.

Improving polygon image segmentation by enhancing U-Net architecture

The crucial task of polyp recognition in medical imaging plays a pivotal role in the early detection and prevention of colorectal cancer. Semantic segmentation, particularly utilizing sophisticated deep learning models such as U-Net, has demonstrated promising results in the realm of polyp segmentation. However, the traditional U-Net structure sometimes grapples with accurately delineating the edges of polyps, which subsequently impacts the overall performance of segmentation. To address this issue, the current study introduces a novel approach by proposing a modified framework of U-Net, equipped with an enhanced edge loss function. This function is designed to ameliorate the accuracy of segmentation within polyp images. The aim is to elevate the model’s capacity to capture intricate details, specifically the edges, which is an area where standard U-Net structures often falter. Experimental outcomes of this study serve to underscore the effectiveness of the proposed approach in accomplishing superior segmentation of edges and improved overall performance in polyp recognition. By successfully tackling the challenges inherent to polyp edge segmentation, the modified U-Net model contributes significantly towards more precise diagnostic systems in the field of medical imaging. Consequently, this research is poised to make a valuable contribution to advancements in the prevention and early detection of colorectal cancer.

Multimodal brain tumor image segmentation based on DenseNet.

A brain tumor magnetic resonance image processing algorithm can help doctors to diagnose and treat the patient's condition, which has important application significance in clinical medicine. This paper proposes a network model based on the combination of U-net and DenseNet to solve the problems of class imbalance in multi-modal brain tumor image segmentation and the loss of effective information features caused by the integration of features in the traditional U-net network. The standard convolution blocks of the coding path and decoding path on the original network are improved to dense blocks, which enhances the transmission of features. The mixed loss function composed of the Binary Cross Entropy Loss function and the Tversky coefficient is used to replace the original single cross-entropy loss, which restrains the influence of irrelevant features on segmentation accuracy. Compared with U-Net, U-Net++, and PA-Net the algorithm in this paper has significantly improved the segmentation accuracy, reaching 0.846, 0.861, and 0.782 respectively in the Dice coefficient index of WT, TC, and ET. The PPV coefficient index has reached 0.849, 0.883, and 0.786 respectively. Compared with the traditional U-net network, the Dice coefficient index of the proposed algorithm exceeds 0.8%, 4.0%, and 1.4%, respectively, and the PPV coefficient index in the tumor core area and tumor enhancement area increases by 3% and 1.2% respectively. The proposed algorithm has the best performance in tumor core area segmentation, and its Sensitivity index has reached 0.924, which has good research significance and application value.

Repurposing traditional U-Net predictions for sparse SAM prompting in medical image segmentation

Objective: Automated medical image segmentation (MIS) using deep learning has traditionally relied on models built and trained from scratch, or at least fine-tuned on a target dataset. The Segment Anything Model (SAM) by Meta challenges this paradigm by providing zero-shot generalisation capabilities. This study aims to develop and compare methods for refining traditional U-Net segmentations by repurposing them for automated SAM prompting. Approach: A 2D U-Net with EfficientNet-B4 encoder was trained using 4-fold cross-validation on an in-house brain metastases dataset. Segmentation predictions from each validation set were used for automatic sparse prompt generation via a bounding box prompting method (BBPM) and novel implementations of the point prompting method (PPM). The PPMs frequently produced poor slice predictions (PSPs) that required identification and substitution. A slice was identified as a PSP if it (1) contained multiple predicted regions per lesion or (2) possessed outlier foreground pixel counts relative to the patient’s other slices. Each PSP was substituted with a corresponding initial U-Net or SAM BBPM prediction. The patients’ mean volumetric dice similarity coefficient (DSC) was used to evaluate and compare the methods’ performances. Main results: Relative to the initial U-Net segmentations, the BBPM improved mean patient DSC by 3.93 ± 1.48% to 0.847 ± 0.008 DSC. PSPs constituted 20.01–21.63% of PPMs’ predictions and without substitution performance dropped by 82.94 ± 3.17% to 0.139 ± 0.023 DSC. Pairing the two PSP identification techniques yielded a sensitivity to PSPs of 92.95 ± 1.20%. By combining this approach with BBPM prediction substitution, the PPMs achieved segmentation accuracies on par with the BBPM, improving mean patient DSC by up to 4.17 ± 1.40% and reaching 0.849 ± 0.007 DSC. Significance: The proposed PSP identification and substitution techniques bridge the gap between PPM and BBPM performance for MIS. Additionally, the uniformity observed in our experiments’ results demonstrates the robustness of SAM to variations in prompting style. These findings can assist in the design of both automatically and manually prompted pipelines.

A Semantic Segmentation Method Based on AS-Unet++ for Power Remote Sensing of Images.

In order to achieve the automatic planning of power transmission lines, a key step is to precisely recognize the feature information of remote sensing images. Considering that the feature information has different depths and the feature distribution is not uniform, a semantic segmentation method based on a new AS-Unet++ is proposed in this paper. First, the atrous spatial pyramid pooling (ASPP) and the squeeze-and-excitation (SE) module are added to traditional Unet, such that the sensing field can be expanded and the important features can be enhanced, which is called AS-Unet. Second, an AS-Unet++ structure is built by using different layers of AS-Unet, such that the feature extraction parts of each layer of AS-Unet are stacked together. Compared with Unet, the proposed AS-Unet++ automatically learns features at different depths and determines a depth with optimal performance. Once the optimal number of network layers is determined, the excess layers can be pruned, which will greatly reduce the number of trained parameters. The experimental results show that the overall recognition accuracy of AS-Unet++ is significantly improved compared to Unet.

Multi-Level Residual Dual Attention Network for Major Cerebral Arteries Segmentation in MRA Toward Diagnosis of Cerebrovascular Disorders.

Segmentation of major brain vessels is very important for the diagnosis of cerebrovascular disorders and subsequent surgical planning. Vessel segmentation is an important preprocessing step for a wide range of algorithms for the automatic diagnosis or treatment of several vascular pathologies and as such, it is valuable to have a well-performing vascular segmentation pipeline. In this article, we propose an end-to-end multiscale residual dual attention deep neural network for resilient major brain vessel segmentation. In the proposed network, the encoder and decoder blocks of the U-Net are replaced with the multi-level atrous residual blocks to enhance the learning capability by increasing the receptive field to extract the various semantic coarse- and fine-grained features. Dual attention block is incorporated in the bottleneck to perform effective multiscale information fusion to obtain detailed structure of blood vessels. The methods were evaluated on the publicly available TubeTK data set. The proposed method outperforms the state-of-the-art techniques with dice of 0.79 on the whole-brain prediction. The statistical and visual assessments indicate that proposed network is robust to outliers and maintains higher consistency in vessel continuity than the traditional U-Net and its variations.

Improved U-net network asphalt pavement crack detection method.

Road crack detection is one of the important parts of road safety detection. Aiming at the problems such as weak segmentation effect of basic U-Net on pavement crack, insufficient precision of crack contour segmentation, difficult to identify narrow crack and low segmentation accuracy, this paper proposes an improved U-net network pavement crack segmentation method. VGG16 and Up_Conv (Upsampling Convolution) modules are introduced as backbone network and feature enhancement network respectively, and the more abstract features in the image are extracted by using the Block depth separable convolution blocks, and the multi-scale features are captured and enhanced by higher level semantic information to improve the recognition accuracy of narrow cracks in the road surface. The improved network embedded the Ca(Channel Attention) attention mechanism in U-net's jump connection to enhance the crack portion to suppress background noise. At the same time, DG_Conv(Depthwise GSConv Convolution) module and UnetUp(Unet Upsampling) module are added in the decoding part to extract richer features through more convolutional layers in the network, so that the model pays more attention to the detailed part of the crack, so the segmentation accuracy can be improved. In order to verify the model's ability to detect cracks in complex backgrounds, experiments were carried out on CFD and Deepcrack datasets. The experimental results show that compared with the traditional U-net network F1-score and mIoU have increased by 13.6% and 9.9% respectively. Superior to advanced models such as U-net, Segnet and Linknet in accuracy and generalization ability, the improved model provides a new method for asphalt pavement crack detection. The model is more conducive to practical application and ground deployment, and can be applied in road maintenance projects.

Segmentation of Seabed Sediment Images Based on Convolutional Neural Network

The analysis of particle sizes in seabed sediments plays an important role in various fields. However, acquiring digital sediment images has been challenging due to the characteristics of sediment particles and the complexity of the seabed exploration environment. The goal of this study is to address the issue of particle segmentation in sediment images and propose an innovative network called SAMU-Net. Innovations are as follows. Firstly, SAMU-Net combines spatial attention modules with the traditional U-Net model, which allows for contextual information extraction and adaptive perception of feature variations. Secondly, the model adopts the idea of multi-scale fusion, enabling networks to better understand the structural and semantic information of the images by fusing feature maps at different scales. In addition, using the drop blocks and normalization, we speed up the training convergence of the model and avoid the overfitting phenomenon. The results show that the segmentation results predicted by the SAMU-Net network are highly similar to the ground truth and have a better segmentation performance compared to the existing algorithms. At the same time, it provides a new solution idea to uneven illumination and other noise interference problems in the digital imaging information of sediments.

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.

Binary semantic segmentation for detection of prostate adenocarcinoma using an ensemble with attention and residual U-Net architectures.

An accurate determination of the Gleason Score (GS) or Gleason Pattern (GP) is crucial in the diagnosis of prostate cancer (PCa) because it is one of the criterion used to guide treatment decisions for prognostic-risk groups. However, the manually designation of GP by a pathologist using a microscope is prone to error and subject to significant inter-observer variability. Deep learning has been used to automatically differentiate GP on digitized slides, aiding pathologists and reducing inter-observer variability, especially in the early GP of cancer. This article presents a binary semantic segmentation for the GP of prostate adenocarcinoma. The segmentation separates benign and malignant tissues, with the malignant class consisting of adenocarcinoma GP3 and GP4 tissues annotated from 50 unique digitized whole slide images (WSIs) of prostate needle core biopsy specimens stained with hematoxylin and eosin. The pyramidal digitized WSIs were extracted into image patches with a size of 256 × 256 pixels at a magnification of 20×. An ensemble approach is proposed combining U-Net-based architectures, including traditional U-Net, attention-based U-Net, and residual attention-based U-Net. This work initially considers a PCa tissue analysis using a combination of attention gate units with residual convolution units. The performance evaluation revealed a mean Intersection-over-Union of 0.79 for the two classes, 0.88 for the benign class, and 0.70 for the malignant class. The proposed method was then used to produce pixel-level segmentation maps of PCa adenocarcinoma tissue slides in the testing set. We developed a screening tool to discriminate between benign and malignant prostate tissue in digitized images of needle biopsy samples using an AI approach. We aimed to identify malignant adenocarcinoma tissues from our own collected, annotated, and organized dataset. Our approach returned the performance which was accepted by the pathologists.

Three-Dimensional Velocity Field Interpolation Based on Attention Mechanism

The establishment of a three-dimensional velocity field is an essential step in seismic exploration, playing a crucial role in understanding complex underground geological structures. Accurate 3D velocity fields are significant for seismic imaging, observation system design, precise positioning of underground geological targets, structural interpretation, and reservoir prediction. Therefore, obtaining an accurate 3D velocity field is a focus and challenge in this field of study. To achieve intelligent interpolation of the 3D velocity field more accurately, we have built a network model based on the attention mechanism, JointA 3DUnet. Based on the traditional U-Net, we have added triple attention blocks and channel attention blocks to enhance dimension information interaction, while adapting to the different changes of geoscience data in horizontal and vertical directions. Moreover, the network also incorporates dilated convolution to enlarge the receptive field. During the training process, we introduced transfer learning to further enhance the network’s performance for interpolation tasks. At the same time, our method is a deep learning interpolation algorithm based on an unsupervised model. It does not require a training set and learns information solely from the input data, automatically interpolating the missing velocity data at the missing positions. We tested our method on both synthetic and real data. The results show that, compared with traditional intelligent interpolation methods, our approach can effectively interpolate the three-dimensional velocity field. The SNR increased to 36.22 dB, and the pointwise relative error decreased to 0.89%.

An improved deep dilated convolutional neural network for seismic facies interpretation

With the successful application and breakthrough of deep learning technology in image segmentation, there has been continuous development in the field of seismic facies interpretation using convolutional neural networks. These intelligent and automated methods significantly reduce manual labor, particularly in the laborious task of manually labeling seismic facies. However, the extensive demand for training data imposes limitations on their wider application. To overcome this challenge, we adopt the UNet architecture as the foundational network structure for seismic facies classification, which has demonstrated effective segmentation results even with small-sample training data. Additionally, we integrate spatial pyramid pooling and dilated convolution modules into the network architecture to enhance the perception of spatial information across a broader range. The seismic facies classification test on the public data from the F3 block verifies the superior performance of our proposed improved network structure in delineating seismic facies boundaries. Comparative analysis against the traditional UNet model reveals that our method achieves more accurate predictive classification results, as evidenced by various evaluation metrics for image segmentation. Obviously, the classification accuracy reaches an impressive 96%. Furthermore, the results of seismic facies classification in the seismic slice dimension provide further confirmation of the superior performance of our proposed method, which accurately defines the range of different seismic facies. This approach holds significant potential for analyzing geological patterns and extracting valuable depositional information.

Multi-parameter Inversion of Tunnels Based on Deep Learning Constraints

When tunnels traverse unfavorable geological formations, they often encounter the risk of sudden water-related incidents, posing a severe threat to the safety of construction workers and property. Consequently, there is an urgent demand for conducting research on the early detection of potential water-related disasters ahead of tunnel excavation. Among the various advanced exploration techniques, the DC resistivity method has proven to be more cost-effective and highly responsive to water-bearing structures, making it a prevalent choice for addressing sudden water hazards in real tunnel projects. This study introduces an innovative electrical approach underpinned by deep learning constraints, fusing resistivity and chargeability to pinpoint the location and size of underground water bodies. More specifically, we’ve designed a learning model rooted in the traditional U-Net convolutional network architecture, featuring a single encoder for feature extraction and a dual decoder for output. When compared to conventional linear inversion techniques and standalone parameter-based deep learning inversion methods, our proposed joint inversion approach exhibits superior performance.

Identification of historic building “genes” based on deep learning: a case study on Chinese baroque architecture in Harbin, China

The monitoring and protection of historic buildings require a highly professional team and material resources. Monitoring and protecting historical architectural features is an urgent issue. According to the theory of biological gene expression, genes are the fundamental units that control and express biological traits. Similarly, the “genes” of historical architecture are the basic units that control historic features. Identifying these historical architecture “genes” involves identifying the main factors that control the historic features. This process is important for monitoring and protecting the historic features. At present, qualitative subjectivity, difficult quantification, poor recognition accuracy, and low reasoning and recognition efficiency exist in the genetic identification of historic buildings. As an example, this article describes Chinese Baroque architecture in Harbin, China, and draws on the principles of biological gene recognition to reference methods of architectural gene recognition in cultural geography and architecture. Improved U-Net models, traditional U-Net models, FCN models, and EfficientNet models that incorporate channel attention mechanisms are used to identify historic building genes, obtaining the optimal intelligent recognition for historical architectural genes based on deep learning. This research shows that the accuracy of an improved U-Net model incorporating a channel attention mechanism is 69%, which is 4%, 7%, and 1% higher than those of the traditional U-Net, FCN, and EfficientNet, respectively. The F1 score of the improved U-Net model reaches 0.654, which is higher than the 0.619 of the traditional U-Net model, 0.645 of the EfficientNet model, and 0.501 of the FCN model. Therefore, the improved U-Net model is the optimal method for identifying historical architecture genes. This research can provide new tools and methods for identifying historical architectural genes.

Semantic Segmentation of High-Resolution Remote Sensing Images with Improved U-Net Based on Transfer Learning

Semantic segmentation of high-resolution remote sensing images has emerged as one of the foci of research in the remote sensing field, which can accurately identify objects on the ground and determine their localization. In contrast, the traditional deep learning-based semantic segmentation, on the other hand, requires a large amount of annotated data, which is unsuitable for high-resolution remote sensing tasks with limited resources. It is therefore important to build a semantic segmentation method for high-resolution remote sensing images. In this paper, it is proposed an improved U-Net model based on transfer learning to solve the semantic segmentation problem of high-resolution remote sensing images. The model is based on the symmetric encoder–decoder structure of U-Net. For the encoder, transfer learning is applied and VGG16 is used as the backbone of the feature extraction network, and in the decoder, after upsampling using bilinear interpolation, it is performed multiscale fusion with the feature maps of the corresponding layers of the encoder in turn and is finally obtained the predicted value of each pixel to achieve precise localization. To verify the efficacy of the proposed network, experiments are performed on the ISPRS Vaihingen dataset. The experiments show that the applied method has achieved high-quality semantic segmentation results on the high-resolution remote sensing dataset, and the MIoU is 1.70%, 2.20%, and 2.33% higher on the training, validation, and test sets, respectively, and the IoU is 4.26%, 6.89%, and 5.44% higher for the automotive category compared to the traditional U-Net.