Segmentation Module Research Articles

Semantic segmentation of substation tools using an improved ICNet network

<p>In the field of substation operation and maintenance, real-time detection and precise segmentation of tools play an important role in maintaining the safe operation of the power grid and guiding operators to work safely. To improve the accuracy and real-time performance of semantic segmentation of substation operation and maintenance tools, we have proposed an improved, light-weight, real-time, semantic segmentation network based on an efficient image cascade network architecture (ICNet). The network uses multiscale branches and cascaded feature fusion units to extract rich multilevel features. We designed a semantic segmentation and purification module to deal with redundant and conflicting information in multiscale feature fusion. A lightweight backbone network was used in the feature extraction stage at different resolutions, and a recursive gated convolution was used in the upsampling stage to achieve high-order spatial interactions, thereby improving segmentation accuracy. Due to the lack of a substation tool semantic segmentation data set, we constructed one. Training and testing on the data set showed that the proposed model improved the accuracy of tool detection while ensuring real-time performance. Compared with the currently popular semantic segmentation network, it had better performance in real-time and accuracy, and provided a new semantic segmentation method for embedded platforms.</p>

Optimized Design of Instrument Recognition Based on CNN Model

Abstract Intelligent recognition of instrument features plays an important role in automation management and overhaul and also facilitates the realization of accurate reading of key parameters in complex environments. The instrument dial intelligent recognition system proposed in this paper consists of geometry correction, pointer segmentation, and reading recognition modules. Combining the idea of the GhostNet model to improve the structure of the backbone network of the Mask RCNN model, the attention mechanism is introduced into the U-Net model, and the minimum outer rectangle method is used for reading recognition. Under different viewpoint rotation angles, the recognition errors of this paper’s method are relatively stable, and they are less than 1%. The region segmentation precision, recall, and accuracy are 99.39%, 99.05%, and 98.38%, respectively. The average error of the recognition results is only -0.04°C, which is satisfactory for instrument recognition.

A real-time interactive restoration system for intraoral digital videos using segment anything model.

Poor conditions in the intraoral environment often lead to low-quality photos and videos, hindering further clinical diagnosis. To restore these digital records, this study proposes a real-time interactive restoration system using segment anything model. Intraoral digital videos, obtained from the vident-lab dataset through an intraoral camera, serve as the input for interactive restoration system. The initial phase employs an interactive segmentation module leveraging segment anything model. Subsequently, a real-time intraframe restoration module and a video enhancement module were designed. A series of ablation studies were systematically conducted to illustrate the superior design of interactive restoration system. Our quantitative evaluation criteria contain restoration quality, segmentation accuracy, and processing speed. Furthermore, the clinical applicability of the processed videos was evaluated by experts. Extensive experiments demonstrated its performance on segmentation with a mean intersection-over-union of 0.977. On video restoration, it leads to reliable performances with peak signal-to-noise ratio of 37.09 and structural similarity index measure of 0.961, respectively. More visualization results are shown on the https://yogurtsam.github.io/iveproject page. Interactive restoration system demonstrates its potential to serve patients and dentists with reliable and controllable intraoral video restoration.

Lymphocyte-Infiltrated Periportal Region Detection With Structurally-Refined Deep Portal Segmentation and Heterogeneous Infiltration Features.

Goal: The early diagnosis and treatment of hepatitis is essential to reduce hepatitis-related liver function deterioration and mortality. One component of the widely-used Ishak grading system for the grading of periportal interface hepatitis is based on the percentage of portal borders infiltrated by lymphocytes. Thus, the accurate detection of lymphocyte-infiltrated periportal regions is critical in the diagnosis of hepatitis. However, the infiltrating lymphocytes usually result in the formation of ambiguous and highly-irregular portal boundaries, and thus identifying the infiltrated portal boundary regions precisely using automated methods is challenging. This study aims to develop a deep-learning-based automatic detection framework to assist diagnosis. Methods: The present study proposes a framework consisting of a Structurally-REfined Deep Portal Segmentation module and an Infiltrated Periportal Region Detection module based on heterogeneous infiltration features to accurately identify the infiltrated periportal regions in liver Whole Slide Images. Results: The proposed method achieves 0.725 in F1-score of lymphocyte-infiltrated periportal region detection. Moreover, the statistics of the ratio of the detected infiltrated portal boundary have high correlation to the Ishak grade (Spearman's correlations more than 0.87 with p-values less than 0.001) and medium correlation to the liver function index aspartate aminotransferase and alanine aminotransferase (Spearman's correlations more than 0.63 and 0.57 with p-values less than 0.001). Conclusions: The study shows the statistics of the ratio of infiltrated portal boundary have correlation to the Ishak grade and liver function index. The proposed framework provides pathologists with a useful and reliable tool for hepatitis diagnosis.

Craniomaxillofacial Bone Segmentation and Landmark Detection Using Semantic Segmentation Networks and an Unbiased Heatmap.

Craniomaxillofacial (CMF) surgery always relies on accurate preoperative planning to assist surgeons, and automatically generating bone structures and digitizing landmarks for CMF preoperative planning is crucial. Since the soft and hard tissues of the CMF regions possess complicated attachment, segmenting the CMF bones and detecting the CMF landmarks are challenging problems. In this study, we proposed a semantic segmentation network to segment the maxilla, mandible, zygoma, zygomatic arch, and frontal bones. Then, we obtained the minimum bounding box around the CMF bones. After cropping, we used the top-down heatmap landmark detection network, similar to the segmentation module, to identify 18 CMF landmarks from the cropping patch. In addition, an unbiased heatmap encoding method was proposed to generate actual landmark coordinates in the heatmap. To overcome quantization effects in the heatmap-based landmark detection networks, the distribution-prior coordinate representation of medical landmarks (DCRML) was proposed to utilize the prior distribution of the encoding heatmap, approximating the accurate landmark coordinates in heatmap decoding by Taylor's theorem. The encoding and decoding method can easily contribute to other existing landmark detection frameworks based on heatmaps; consequently, these approaches can readily benefit without changing model structure. We used prior segmentation knowledge to enhance the semantic information around the landmarks, increasing landmark detection accuracy. The proposed framework was evaluated by 100 healthy persons and 86 patients from multicenter cooperation. The mean Dice score of our proposed segmentation network achieved over 88 %; in particular, the mandible accuracy was approximately 95%. The mean error of landmarks was 1.84 ±1.32 mm.

RPFNet: Recurrent Pyramid Frequency Feature Fusion Network for Instance Segmentation in Side-Scan Sonar Images

Side-scan sonar (SSS) is an essential acoustic sensor device for obtaining underwater information. The instance segmentation of sonar images can effectively locate and detect underwater objects. Although various CNN-based frameworks have achieved promising results in natural image instance segmentation, the noise interference, highlight shadow, and blurred edge in sonar images bring more significant challenges for sonar image instance segmentation. To solve these problems, we propose a novel recurrent pyramid frequency feature fusion network (RPFNet), which mainly consists of the pyramid frequency feature fusion network (PFN), recurrent residual attention mechanism (RRAM), mask prediction module, and semantic segmentation module. By enhancing and fusing different frequency features of SSS images, PFN can effectively extract fine-grained features and reduce background information interference. The RRAM uses residual structure and attention mechanism to enhance the correlation of different frequency features and improve nonlinear feature representation ability. The mask prediction and semantic segmentation modules are used to discriminate object instance categories and generate corresponding semantic segmentation results. We extensively evaluate the proposed framework on the real-scenes sonar image dataset. The evaluation results demonstrate that the proposed method outperforms the state-of-the-art methods in various evaluation metrics. Code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/darkseid-arch/SonarRPFNet</uri> .

Dense Top-View Semantic Completion With Sparse Guidance and Online Distillation

Dense semantic scene understanding of the surrounding environment in top-view is a crucial task for autonomous vehicles. Recent LiDAR-based semantic perception works mainly focus on point-wise predictions of the LiDAR points instead of dense predictions of the environment, making them not appropriate for path-planning tasks. Pillar and voxel representations can achieve dense predictions, but the generation of data representation and data processing are usually time-consuming. In this paper, we propose a top-view semantic completion network to produce accurate dense grid-wise predictions with real-time performance. Specifically, we propose an online distillation strategy, consisting of two parts: a student model using 2D range-view and top-view representations, and a teacher model using range-view, top-view, and voxel representations. To realize information transfer between different representations, we propose a cross-view association (CVA) module, by which the range-view features and 3D voxel features are converted into the ones in the top-view. The proposed method can avoid the difficulty of direct dense semantic segmentation in the top-view, with the point-wise sparse semantic segmentation module acting as a guide for the dense grid-wise semantic completion in a semantic-completion way. It can also alleviate the computational complexity by using only the voxel representation and 3D convolution in the teacher model. The experimental results on the SemanticKITTI dataset (46.4% mIoU) and nuScenes-LidarSeg dataset (47.3% mIoU) demonstrate the effectiveness of the proposed sparse guidance and online distillation strategies.

YOLOv8-segANDcal: segmentation, extraction, and calculation of soybean radicle features.

The high-throughput and full-time acquisition of images of crop growth processes, and the analysis of the morphological parameters of their features, is the foundation for achieving fast breeding technology, thereby accelerating the exploration of germplasm resources and variety selection by crop breeders. The evolution of embryonic soybean radicle characteristics during germination is an important indicator of soybean seed vitality, which directly affects the subsequent growth process and yield of soybeans. In order to address the time-consuming and labor-intensive manual measurement of embryonic radicle characteristics, as well as the issue of large errors, this paper utilizes continuous time-series crop growth vitality monitoring system to collect full-time sequence images of soybean germination. By introducing the attention mechanism SegNext_Attention, improving the Segment module, and adding the CAL module, a YOLOv8-segANDcal model for the segmentation and extraction of soybean embryonic radicle features and radicle length calculation was constructed. Compared to the YOLOv8-seg model, the model respectively improved the detection and segmentation of embryonic radicles by 2% and 1% in mAP50-95, and calculated the contour features and radicle length of the embryonic radicles, obtaining the morphological evolution of the embryonic radicle contour features over germination time. This model provides a rapid and accurate method for crop breeders and agronomists to select crop varieties.

Mixed noise-guided mutual constraint framework for unsupervised anomaly detection in smart industries

Large-scale sensor and data acquisition systems, integrated with deep learning methodologies, play a pivotal role in enhancing the sustainability and security of smart city environments, exemplifying the critical significance of anomaly detection techniques. Anomaly detection in complex industrial scenarios presents various challenges, such as intricate working environments, limited anomaly samples, and lack of a priori information. Unsupervised anomaly detection based on knowledge distillation enables anomaly detection using only normal samples. However, the similarity in structure between teacher and student models, along with identical input data flow, hampers accurate anomaly detection and localization. To address these issues, we propose MNMC, an unsupervised anomaly detection model consisting of a mixed noise generation module emulating real defects, a mutual constraint module, and an anomaly segmentation module. Firstly, to enhance the student network’s ability to learn robust features, we construct a hybrid noise model comprising dead-leaves noise and perlin noise. This generates features with structural texture and distributional characteristics closer to real anomalies. Secondly, we design a mutual constraint framework to further improve the learning ability of the student network for normal features by constraining representations containing only a single noise. Lastly, for the detection of anomalies at different scales, we propose a new evaluation metric based on equal importance of normal and anomalous regions. Through ablation experiments, we demonstrate the effectiveness of the simulated real defect generation module and the mutual constraints module. Performance experiments on the MVTec dataset show that our method achieves competitive results compared to the current state-of-the-art anomaly detection methods.

Actively Reconfigurable Segmented Spatial Sound Modulators

AbstractHigh‐quality acousto‐holographic patterns and images, integral to applications like 3D displays, acoustophoresis, and midair haptics, require precise distribution of ultrasound waves to achieve. Essential tools for this task are spatial sound modulators (SSMs), which control constituent elements to enable dynamic distribution of sound pressure. However, current ultrasonic SSMs face limitations due to high costs and the intricate actuation of numerous small, closely spaced units. This study introduces “segmented SSMs,” novel devices that combine traditional acoustic metasurface pixel units into custom‐shaped segmented elements. These segmented SSMs reduce actuation costs and complexity while retaining pressure distribution quality. This approach includes a custom phase agglomeration algorithm (PAA), that offers a hierarchy of potential segmentation solutions for user selection. An SSM fabrication method is detailed using off‐the‐shelf 3D printers and bespoke control electronics, completing an end‐to‐end methodology from conception to realization. This approach is validated with two prototype SSM devices that focus sound waves and levitate polystyrene beads using dynamic segmented elements. Further enhancements to the technique are explored through hybrid SSM devices with both static and dynamic elements. The pipeline facilitates efficient SSM construction across diverse applications and invites the inception of future devices with varying sizes, uses, and actuation mechanisms.

PCNN orchard heterologous image fusion with semantic segmentation of significance regions

The apple fruits captured from red green and blue (RGB) color camera cannot be accurately identified and localized under natural scenes of orchards due to lighting and shading. We propose a pulse coupled neural network (PCNN) orchard heterogeneous image fusion model for semantic segmentation of saliency regions. The purpose of the model is to improve the accuracy of image segmentation and the quality of heterogeneous source image fusion in complex orchard scenes. The model fuses the time of flight (ToF) and RGB images acquired from orchard scenes. This model consists of an orchard image semantic segmentation module, a parameter optimization module, and a dual PCNN image fusion module. Image semantic segmentation module extracts the significant regions of fruit targets in orchard images. And then, in parameter optimization module, we introduce tent chaos and ranking strategies into human mental search algorithm to optimize the parameters of PCNN model. Finally, we complete the fusion through a dual PCNN module. Six fusion algorithms were used to evaluate the fusion quality of public and self-built datasets. The results showed that the chaotic hierarchical mental search algorithm could better solve the problem of difficult-to-determine PCNN parameters. The improved DeeplabV3 + could also segment clearer and more complete apple regions. Compared with other six fusion algorithms, the fusion experimental results show that the fused images with clearer targets and more prominent textures could be obtained in the proposed model. The proposed model reduces the influence of light and other disturbing factors in fruit recognition. And the fused image contains rich and accurate image information, which is more in line with the perception habit of human eyes.

Adaptive2Former: Enhancing Chromosome Instance Segmentation with Adaptive Query Decoder

Chromosome instance segmentation plays a crucial role in chromosomal karyotype analysis. However, the overlapping of chromosome instances and their individual morphological differences make accurate chromosome instance segmentation a challenging task. Especially in handling overlapping chromosome instances, traditional segmentation methods tend to confuse instances with one another. To solve these problems, this paper proposes an innovative method named Adaptive2Former. It builds upon our novel devised Adaptive Query Decoder (AQD) module to enhance segmentation precision. The AQD effectively utilizes the [ cls ] token from the backbone network to dynamically generate adaptive query vectors instead of using fixed queries. This new design leverages the semantic information inside the input images, producing representations more conducive to subsequent segmentation module, thereby improving the model’s segmentation performance. Experiments conducted on our dataset demonstrate that the proposed Adaptive2Former significantly enhances the performance of chromosome instance segmentation compared to Mask2Former and other existing models, achieving results of 97.65% mAP75 and 0.21 Dice Loss.

DFBU-Net: Double-branch flat bottom U-Net for efficient medical image segmentation

In the field of medical image processing, segmenting tissues and organs in CT/MRI and other medical sequence images is a vital yet challenging task. Analyzing the MICCAI competition, we have identified two problems in current methods for medical image organ segmentation: (1) There is a bottleneck in organ segmentation, with marginal room for improvement, as algorithmic capabilities have already surpassed the task’s inherent difficulty. (2) Most current research focuses on stacking and enhancing new modules for segmentation while overlooking the inherent characteristics of medical sequence images. To overcome these two problems, firstly, we have encapsulated the three characteristics of CT/MRI medical sequence image segmentation: semantic correctness, edge accuracy, and 3D structure. Secondly, we delved into the most information-rich downsampling stage in terms of detail and semantics. Subsequently, we designed a flat-bottom double-branch network (DFBU-Net) based on the U-Net architecture. The high-resolution flat bottom branch of this network maintained a 1/4 feature map size to ensure the preservation of rich detail information, while the low-resolution branch underwent progressive downsampling to capture more semantic information. To prevent information loss, cross-fusion was performed at each stage of the model’s two branches. Finally, DFBU-Net was evaluated on the MICCAI FLARE2021 dataset (DSC:93.61%, NSD:85.01%). Particularly, in the challenging task of pancreatic segmentation, our model outperformed the first-place model by 0.72% in DSC and 2.92% in NSD. Furthermore, in the MICCAI PARSE2022 competition, DFBU-Net ranked ninth with a DICE score of 79.28%, demonstrating its excellent segmentation performance and generalization ability.

Using Spontaneous Speech Recognition as a Biomarker to Distinguish Dementia Patients

AbstractBackgroundDigital biomarkers such as individual speech recordings, are a cost‐effective tool for assessing cognitive status using linguistic and acoustic content. The purpose of this study is to investigate whether Natural Language Processing (NLP) methods and audio feature extraction can distinguish spontaneous speech outputs in dementia patients from a healthy population.Method240 audio samples and transcripts of clinician‐participant dialogues involving the Cookie‐Theft visual description were acquired from the DementiaBank Pitt corpus (Dementia group (DG) n = 169; healthy controls (CG) n = 71). Lexical variety, syntactic complexity, salience of information, disfluency patterns, and sentiment analysis were evaluated. Sound synthesis and an acoustic R analysis package were used to filter and examine acoustic properties of recorded audio such as silence and speech segmentation, pitch tracking, and frequency modulation of vibrato or jitter. DG and CG were compared using non‐parametric Kruskal‐Wallis tests, Dunn’s multiple comparison, and Mann‐Whitney U‐test in the statistical analysis. In comparison to CG, we hypothesized that DG would differ in levels of lexical diversity, grammatical complexity, and overall salience.ResultOverall, text analysis showed no variations in disfluencies between groups, but there were significant differences in lexical diversity, syntactic complexity and reference rate to reality (p&lt;0.0001). Additionally, there were no appreciable differences in the frequency of salient information between the two groups, however the frequency of non‐salient information and disfluencies was considerably higher in the DG (p&lt;0.05). Analysis of acoustic features revealed that DG had longer average speech duration and speech segment durations compared to CG (p&lt;0.001).. On the other hand, the rate of change of acoustic characteristics including pitch, mean entropy, and harmonics‐to‐noise ratios did not significantly differ between DG and CG.ConclusionIn conclusion, this paradigm reveals that, in comparison to healthy controls, dementia patients exhibited increased lexical diversity yet lower linguistic complexity, and a shift towards characterizing non‐salient objects in given stimuli, as well as longer segments of speech. Further examination of the rhythmic and acoustic characteristics of speech may provide a comprehensive framework towards defining speech as a potential digital biomarker for dementia risk, and provide applications to regions experiencing increasing socioeconomic disparities.

Development and Clinical Application of Artificial Intelligence Assistant System for Rotator Cuff Ultrasound Scanning

We developed an intelligent assistance system for shoulder ultrasound imaging, incorporating deep-learning algorithms to facilitate standard plane recognition and automatic tissue segmentation of the rotator cuff and its surrounding structures. We evaluated the system's performance using a dedicated data set of rotator cuff ultrasound images to assess its feasibility in clinical practice. To fulfill the system's primary functions, we designed a standard plane recognition module based on the ResNet50 network and an automatic tissue segmentation module using the Mask R-CNN model. The modules were trained on carefully curated data sets. The standard plane recognition module automatically identifies a specific standard plane based on the ultrasound image characteristics. The automatic tissue segmentation module effectively delineates and segments anatomical structures within the identified standard plane. With the use of 59,265 shoulder joint ultrasound images, the standard plane recognition model achieved an impressive recognition accuracy of 94.9% in the test set, with an average precision rate of 96.4%, recall rate of 95.4% and F1 score of 95.9%. The automatic tissue segmentation model, tested on 1886 images, exhibited a commendable average intersection over union value of 96.2%, indicating robustness and accuracy. The model achieved mean intersection over union values exceeding 90.0% for all standard planes, indicating its effectiveness in precisely delineating the anatomical structures. Our shoulder joint musculoskeletal intelligence system swiftly and accurately identifies standard planes and performs automatic tissue segmentation.

TSRNet: Tongue image segmentation with global and local refinement

Tongue Image Segmentation is an essential task for intelligent Traditional Chinese Medicine (TCM), as the tongue is sensitive to the physiological conditions and pathological changes of patients and can help physicians determine strategies for the syndrome differentiation. However, it is a big challenge to acquire an accurate tongue segmentation mask, due to the varied shape and texture of the tongue. This paper proposes a novel tongue segmentation network based on an encoder–decoder framework with global and local refinement, named TSRNet. In the global refinement module, we design an effective module for fusing features from an autoencoder, which is pre-trained on tongue images with segmentation labels, so that the network can make use of the prior knowledge. Moreover, in the local refinement module, we perform patch sampling according to the coarse prediction boundary and correct errors through a patch segmentation module. Both two modules are plugged into the decoder to obtain better tongue segmentation results by training end-to-end. Experimental results compared with state-of-the-art models on two real-world tongue datasets demonstrate the effectiveness of the proposed TSRNet.

An efficient point cloud place recognition approach based on transformer in dynamic environment

Place recognition retrieves scenes from a prior map and identifies previously visited locations. It is essential to autonomous-driving and robotics in terms of long-term navigation and global localization. Although the semantic graph has made great progress in 3D place recognition, its construction is greatly affected by the complex dynamic environment. In response, this paper designs a scene graph transformation network to solve place recognition in a dynamic environment. To remove the interference of moving objects, we introduce a moving object segmentation (MOS) module. Besides, we design a scene graph transformer-attention module to generate a more discriminative and representative global scene graph descriptor, which significantly improves the performance of place recognition. In addition, we integrate our place recognition method for loop closure with an existing LiDAR-based odometry, boosting its localization accuracy. We evaluate our method on the KITTI and Oxford RobotCar datasets. Many experimental results show that our method can effectively accomplish place recognition, and its accuracy and robustness improve by at least 3% when compared with existing state-of-the-art methods, such as DiSCO. To illustrate the generalization capabilities of our method, we evaluate it on the KITTI-360 and NCLT datasets while using only KITTI for training. The experiments show that our scene graph descriptor can achieve accurate loop-closure and global localization in never-seen environments.

An Advanced Software Platform and Algorithmic Framework for Mobile DBH Data Acquisition

Rapid and precise tree Diameter at Breast Height (DBH) measurement is pivotal in forest inventories. While the recent advancements in LiDAR and Structure from Motion (SFM) technologies have paved the way for automated DBH measurements, the significant equipment costs and the complexity of operational procedures continue to constrain the ubiquitous adoption of these technologies for real-time DBH assessments. In this research, we introduce KAN-Forest, a real-time DBH measurement and key point localization algorithm utilizing RGB-D (Red, Green, Blue-Depth) imaging technology. Firstly, we improved the YOLOv5-seg segmentation module with a Channel and Spatial Attention (CBAM) module, augmenting its efficiency in extracting the tree’s edge features in intricate forest scenarios. Subsequently, we devised an image processing algorithm for real-time key point localization and DBH measurement, leveraging historical data to fine-tune current frame assessments. This system facilitates real-time image data upload via wireless LAN for immediate host computer processing. We validated our approach on seven sample plots, achieving bbAP50 and segAP50 scores of: 90.0%(+3.0%), 90.9%(+0.9%), respectively with the improved YOLOv5-seg model. The method exhibited a DBH estimation RMSE of 17.61∼54.96 mm (R2=0.937), and secured 78% valid DBH samples at a 59 FPS. Our system stands as a cost-effective, portable, and user-friendly alternative to conventional forest survey techniques, maintaining accuracy in real-time measurements compared to SFM- and LiDAR-based algorithms. The integration of WLAN and its inherent scalability facilitates deployment on Unmanned Ground Vehicles (UGVs) to improve the efficiency of forest inventory. We have shared the algorithms and datasets on Github for peer evaluations.

Reversed domain adaptation for nuclei segmentation-based pathological image classification

Despite the fact that digital pathology has provided a new paradigm for modern medicine, the insufficiency of annotations for training remains a significant challenge. Due to the weak generalization abilities of deep-learning models, their performance is notably constrained in domains without sufficient annotations. Our research aims to enhance the model’s generalization ability through domain adaptation, increasing the prediction ability for the target domain data while only using the source domain labels for training. To further enhance classification performance, we introduce nuclei segmentation to provide the classifier with more diagnostically valuable nuclei information. In contrast to the general domain adaptation that generates source-like results in the target domain, we propose a reversed domain adaptation strategy that generates target-like results in the source domain, enabling the classification model to be more robust to inaccurate segmentation results. The proposed reversed unsupervised domain adaptation can effectively reduce the disparities in nuclei segmentation between the source and target domains without any target domain labels, leading to improved image classification performance in the target domain. The whole framework is designed in a unified manner so that the segmentation and classification modules can be trained jointly. Extensive experiments demonstrate that the proposed method significantly improves the classification performance in the target domain and outperforms existing general domain adaptation methods.

MDSSN: An end-to-end deep network on triangle mesh parameterization

Mesh data plays a crucial role in a wide range of applications worldwide within the field of 3D computer vision. In contrast to traditional Euclidean space arrangements, meshes encompass spatial information, including edges, faces, angles, and graph structures that embed face information within the mesh data. Nevertheless, conventional deep learning frameworks such as convolutional neural networks (CNNs) have struggled to achieve significant advancements in handling meshes. This paper proposes a simple mesh computation framework called Mesh Decomposition Second-order Sobolev Network (MDSSN) to model triangle meshes and represent their shape. The construction of face-based Riemannian and edge-based Riemannian graphs is motivated by the remarkable representation capabilities of edges and faces in mesh data, achieved through a graph composition mechanism. Furthermore, we design end-to-end operators, including the Sobolev-filter block, pooling, and unpooling blocks, which draw inspiration from traditional deep learning frameworks for modeling the mesh structure. Importantly, the design of the mesh-filter block incorporates the second-order Sobolev metric. To address common challenges in mesh classification and segmentation tasks, we have designed dedicated classification and segmentation modules. We utilize a joint loss function that integrates losses from adjacent faces and categories. We evaluate the performance of the MDSSN in mesh classification and segmentation tasks, focusing on its ability to represent 3D shapes. Experimental results demonstrate that MDSSN achieves superior performance compared to other state-of-the-art approaches.