TIPARP as a prognostic biomarker and potential immunotherapeutic target in male papillary thyroid carcinoma.

BACKGROUND: Male patients with papillary thyroid carcinoma (PTC) tend to have poorer prognosis compared to females, partially attributable to a higher rate of lymph node metastasis (LNM). Developing a precise predictive model for LNM occurrence in male PTC patients is imperative. While preliminary predictive models exist, there is room to improve accuracy. Further research is needed to create optimized prognostic models specific to LNM prediction in male PTC cases. METHODS: We conducted a comprehensive search of publicly available microarray datasets to identify candidate genes continuously upregulated or downregulated during PTC progression in male patients only. Univariate Cox analysis and lasso regression were utilized to construct an 11-gene signature predictive of LNM. TIPARP emerged as a key candidate gene, which we validated at the protein level using immunohistochemical staining. A prognostic nomogram incorporating the signature and clinical factors was developed based on the TCGA cohort. RESULTS: The 11-gene signature demonstrated good discriminative performance for LNM prediction in training and validation datasets. High TIPARP expression associated with advanced stage, high T stage, and presence of LNM. A prognostic nomogram integrating the signature and clinical variables reliably stratified male PTC patients into high and low recurrence risk groups. CONCLUSIONS: We identified a robust 11-gene signature and prognostic nomogram for predicting LNM occurrence in male PTC patients. We propose TIPARP as a potential contributor to inferior outcomes in males, warranting further exploration as a prognostic biomarker and immunotherapeutic target. Our study provides insights into the molecular basis for gender disparities in PTC.

METTL3 (Methyltransferase Like 3)-Dependent N6-Methyladenosine Modification on Braf mRNA Promotes Macrophage Inflammatory Response and Atherosclerosis in Mice.

BACKGROUND: Atherosclerosis is a chronic inflammatory disease, in which macrophages determine the progression of atherosclerotic plaques. However, no studies have investigated how METTL3 (methyltransferase like 3) in macrophages affects atherosclerotic plaque formation in vivo. Additionally, whether Braf mRNA is modified by METTL3-dependent N6-methyladenosine (m6A) methylation remains unknown. METHODS: We analyzed single-cell sequencing data of atherosclerotic plaques in mice fed with a high fat diet for different periods. Mettl3fl/fl Lyz2cre Apoe-/- mice and littermate control Mettl3fl/fl Apoe-/- mice were generated and fed high fat diet for 14 weeks. In vitro, we stimulated peritoneal macrophages with ox-LDL (oxidized low-density lipoprotein) and tested the mRNA and protein expression levels of inflammatory factors and molecules regulating ERK (extracellular signal-regulated kinase) phosphorylation. To find METTL3 targets in macrophages, we performed m6A-methylated RNA immunoprecipitation sequencing and m6A-methylated RNA immunoprecipitation-qPCR. Further, point mutation experiments were used to explore m6A-methylated adenine. Using RNA immunoprecipitation assay, we explored m6A methylation-writing protein bound to Braf mRNA. RESULTS: In vivo, METTL3 expression in macrophages increased with the progression of atherosclerosis. Myeloid cell-specific METTL3 deletion negatively regulated atherosclerosis progression and the inflammatory response. In vitro, METTL3 knockdown or knockout in macrophages attenuated ox-LDL-mediated ERK phosphorylation rather than JNK (c-Jun N-terminal kinase) and p38 phosphorylation and reduced the level of inflammatory factors by affecting BRAF protein expression. The negative regulation of inflammation response caused by METTL3 knockout was rescued by overexpression of BRAF. In mechanism, METTL3 targeted adenine (39725126 in chromosome 6) on the Braf mRNA. Then, YTHDF1 could bind to m6A-methylated Braf mRNA and promoted its translation. CONCLUSIONS: Myeloid cell-specific Mettl3 deficiency suppressed hyperlipidemia-induced atherosclerotic plaque formation and attenuated atherosclerotic inflammation. We identified Braf mRNA as a novel target of METTL3 in the activation of the ox-LDL-induced ERK pathway and inflammatory response in macrophages. METTL3 may represent a potential target for the treatment of atherosclerosis.

A tripartite evolutionary game study of low-carbon innovation system from the perspective of dynamic subsidies and taxes.

Traditional manufacturing industry is in the early stages of transition to low-carbon innovative production, and is in urgent need of a low-carbon innovation system to achieve the goal of carbon neutrality. In order to realize the effective supervision of enterprise carbon emissions, this paper constructs a tripartite evolutionary game model among the corporate, government and public from the perspective of dynamic subsidies and taxes. The main results are as follows. First, the increase in government subsidies to a certain extent will help encourage companies to choose low-carbon innovative production strategies, but more subsidies are not always better. Excessive subsidies will increase the cost of government regulation and reduce the probability of government regulation. Second, the tripartite evolutionary game system does not converge under the static subsidies and taxes mechanism. But the system could quickly converges to the stable condition under dynamic subsidies and taxes. The stable point is the situation of corporate low-carbon innovation, government regulation, and public supervision. Third, the public intervention and supervision can effectively prevent the phenomenon of government misconduct and enterprises over-emission production. And the influence of public reward and punishment is more effective for the government than for enterprises.

O-linked ß-N-acetylglucosamine transferase plays an essential role in heart development through regulating angiopoietin-1.

O-linked N-acetylglucosamine (GlcNAc) transferase (OGT) is the only enzyme catalyzing O-GlcNAcylation. Although it has been shown that OGT plays an essential role in maintaining postnatal heart function, its role in heart development remains unknown. Here we showed that loss of OGT in early fetal cardiomyocytes led to multiple heart developmental defects including hypertrabeculation, biventricular dilation, atrial septal defects, ventricular septal defects, and defects in coronary vessel development. In addition, RNA sequencing revealed that Angiopoietin-1, required within cardiomyocytes for both myocardial and coronary vessel development, was dramatically downregulated in cardiomyocyte-specific OGT knockout mouse hearts. In conclusion, our data demonstrated that OGT plays an essential role in regulating heart development through activating expression of cardiomyocyte Angiopoietin-1.

Opinion cascade under perception bias in social networks.

Opinion cascades, initiated by active opinions, offer a valuable avenue for exploring the dynamics of consensus and disagreement formation. Nevertheless, the impact of biased perceptions on opinion cascade, arising from the balance between global information and locally accessible information within network neighborhoods, whether intentionally or unintentionally, has received limited attention. In this study, we introduce a threshold model to simulate the opinion cascade process within social networks. Our findings reveal that consensus emerges only when the collective stubbornness of the population falls below a critical threshold. Additionally, as stubbornness decreases, we observe a higher prevalence of first-order and second-order phase transitions between consensus and disagreement. The emergence of disagreement can be attributed to the formation of echo chambers, which are tightly knit communities where agents' biased perceptions of active opinions are lower than their stubbornness, thus hindering the erosion of active opinions. This research establishes a valuable framework for investigating the relationship between perception bias and opinion formation, providing insights into addressing disagreement in the presence of biased information.

Online education isn't the best choice: evidence-based medical education in the post-epidemic era-a cross-sectional study.

BACKGROUND: The COVID-19 pandemic led many educational institutions to shift to online courses, making blended education a significant trend in teaching. We examined the effectiveness of blended learning in an evidence-based medicine course. METHODS: We compared the examination scores of a blended learning group, an online only group, and a traditional offline group and conducted a questionnaire survey on students' preferences for different learning modes and the reasons for their preferences. A total of 2100 undergraduate students in clinical medicine were included in this cross-sectional study. Examination results were collected, and questionnaires were administered to the study participants. We compared the mean scores and exam pass rates of the three teaching groups using ANOVA and c2test for multiple comparisons. RESULTS: The blended group's exam scores and pass rate were significantly higher than those of the offline and online groups. Furthermore, 71.6% preferred the blended teaching mode. In the survey on " learning effectiveness", the majority of the students believed that blended education could better enhance the initiative of learning, the interest of the course, the pertinence of the learning content, the comprehension of evidence-based medical thinking, and the basic skills of evidence-based practice. Subsequently, in a questionnaire administered to a blended group of students, their foremost reason for liking online instruction was 'flexible in time and space' (99%), followed by 'can be viewed repeatedly, facilitating a better understanding of knowledge points' (98%). Their foremost reason for liking offline teaching was 'helps to create a good learning atmosphere' (97%), followed by 'teachers can control students' learning status in real time' (89%). CONCLUSIONS: This study explored the effectiveness of learning in evidence-based medicine courses by comparing the learning outcomes and personal perceptions of three different teaching modes. This is the first cross-sectional study in which three different teaching models are compared and discussed in an evidence-based medicine course. We also elaborate on the specific instructional protocols for each model. This study shows that using a blended education approach in evidence-based medicine courses can improve students' learning motivation, autonomy, and satisfaction. It also enhances instructional efficiency, thereby improving students' understanding of the course content.

Single-Pixel Background Modeling Algorithm for Strong Sky Scenes Based on Local Region Spatial Bases.

In the dim-small target detection field, background suppression is a key technique for stably extracting the target. In order to effectively suppress the background to enhance the target, this paper presents a novel background modeling algorithm, which constructs base functions for each pixel based on the local region background and models the background of each pixel, named single pixel background modeling (SPB). In SPB, the low-rank blocks of the local backgrounds are first obtained to construct the background base functions of the center pixel. Then, the background of the center pixel is optimally estimated by the background bases. Experiments demonstrate that in the case of extremely low signal-to-noise ratio (SNR < 1.5 dB) and complex motion state of targets, SPB can stably and effectively separate the target from the strongly undulant sky background. The difference image obtained via SPB background modeling has the characters: the non-target residual could be white noise, and the target is significantly enhanced. Compared with the other typical five algorithms, SPB remarkably outperforms other algorithms to detect the target of a low signal-to-noise ratio.

Impact of Mandibular Angle Osteotomy Using a Geometric Mathematical Design on the Aesthetic Osteotomy Line: A Retrospective Observational Study.

BACKGROUND: Mandibular angle osteotomy (MAO) is a frequently described technique in Eastern females. The success hinges on the precise positioning of the osteotomy line. The geometric mathematical method is viable. Therefore, we explored the impact of mandibular angle osteotomy using aesthetic standards and printed digital osteotomy templates (DOTs) on the aesthetic osteotomy line. METHODS: This retrospective observational study included female patients with prominent mandibular angle (PMA) who underwent MAO at our hospital between January 2020 and March 2021. Thirty-three female patients were included, 22 in the DOTs group using new DOTs, and 11 in the traditional group using traditional free-hand techniques. RESULTS: Regarding the width of the excised bone, the postoperative deviation from the preoperative plan was not significant in the DOTs group (0.5 ± 0.3 mm, P > 0.05), while the deviation was significant for the traditional group (2.5 ± 1.2 mm, P<0.05). The preparation time was longer in the DOTs group than in the traditional group (82 ± 11 vs. 53±4 min, P < 0.001). The osteotomy time and the operation time were shorter in the DOTs group than in the traditional group (osteotomy: 54 ± 5 vs. 73 ± 6 min; preparation: 124 ± 10 vs. 169 ± 13 min; both P < 0.001). The Likert (4.0 ± 0.5 vs. 1.0 ± 0.6, P = 0.006) and FACE-Q scores (17.5 ± 1.7 vs. 15.6 ± 1.3, P = 0.029) were higher in the DOTs group. CONCLUSIONS: The new method of positioning the new aesthetic osteotomy line based on geometric analysis might provide a possible osteotomy method that strongly suggests effectiveness, safety, individualization, and accuracy, with a shorter operation and higher patient satisfaction. LEVEL OF EVIDENCE IV: This journal requires that authors assign a level of evidence to each article. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors www.springer.com/00266 .

Ferritin: A Multifunctional Nanoplatform for Biological Detection, Imaging Diagnosis, and Drug Delivery.

Ferritins are spherical iron storage proteins within cells that are composed of a combination of 24 subunits of two types, heavy-chain ferritin (HFn) and light-chain ferritin (LFn). They autoassemble naturally into a spherical hollow nanocage with an outer diameter of 12 nm and an interior cavity that is 8 nm in diameter. In recent years, with the constantly emerging safety issues and the concerns about unfavorable uniformity and indefinite in vivo behavior of traditional nanomedicines, the characteristics of native ferritin nanocages, such as the unique nanocage structure, excellent safety profile, and definite in vivo behavior, make ferritin-based formulations uniquely attractive for nanomedicine development. To date, a variety of cargo molecules, including therapeutic drugs (e.g., cisplatin, carboplatin, paclitaxel, curcumin, atropine, quercetin, gefitinib, daunomycin, epirubicin, doxorubicin, etc.), imaging agents (e.g., fluorescence dyes, radioisotopes, and MRI contrast agents), nucleic acids (e.g., siRNA and miRNA), and metal nanoparticles (e.g., Fe3O4, CeO2, AuPd, CuS, CoPt, FeCo, Ag, etc.) have been loaded into the interior cavity of ferritin nanocages for a broad range of biomedical applications from in vitro biosensing to targeted delivery of cargo molecules in living systems with the aid of modified targeting ligands either genetically or chemically. We reported that human HFn could selectively deliver a large amount of cargo into tumors in vivo via transferrin receptor 1 (TfR1)-mediated tumor-cell-specific targeting followed by rapid internalization. By the use of the intrinsic tumor-targeting property and unique nanocage structure of human HFn, a broad variety of cargo-loaded HFn formulations have been developed for biological analysis, imaging diagnosis, and medicine development. In view of the intrinsic tumor-targeting property, unique nanocage structure, lack of immunogenicity, and definite in vivo behavior, human HFn holds promise to promote therapeutic drugs, diagnostic imaging agents, and targeting moieties into multifunctional nanomedicines.Since the report of the intrinsic tumor-targeting property of human HFn, we have extensively explored human HFn as an ideal nanocarrier for tumor-targeted delivery of anticancer drugs, MRI contrast agents, inorganic nanoparticles, and radioisotopes. In particular, by the use of genetic tools, we also have genetically engineered human HFn nanocages to recognize a broader range of disease biomarkers. In this Account, we systematically review human ferritins from characterizing their tumor-binding property and understanding their mechanism and kinetics for cargo loading to exploring their biomedical applications. We finally discuss the prospect of ferritin-based formulations to become next-generation nanomedicines. We expect that ferritin formulations with unique physicochemical characteristics and intrinsic tumor-targeting property will attract broad interest in fundamental drug research and offer new opportunities for nanomedicine development.

Cardiomyocyte Expression of ZO-1 Is Essential for Normal Atrioventricular Conduction but Does Not Alter Ventricular Function.

RATIONALE: ZO-1 (Zonula occludens-1), a plasma membrane-associated scaffolding protein regulates signal transduction, transcription, and cellular communication. Global deletion of ZO-1 in the mouse is lethal by embryonic day 11.5. The function of ZO-1 in cardiac myocytes (CM) is largely unknown. OBJECTIVE: To determine the function of CM ZO-1 in the intact heart, given its binding to other CM proteins that have been shown instrumental in normal cardiac conduction and function. METHODS AND RESULTS: We generated ZO-1 CM-specific knockout (KO) mice using α-Myosin Heavy Chain-nuclear Cre (ZO-1cKO) and investigated physiological and electrophysiological function by echocardiography, surface ECG and conscious telemetry, intracardiac electrograms and pacing, and optical mapping studies. ZO-1cKO mice were viable, had normal Mendelian ratios, and had a normal lifespan. Ventricular morphometry and function were not significantly different between the ZO-1cKO versus control (CTL) mice, basally in young or aged mice, or even when hearts were subjected to hemodynamic loading. Atrial mass was increased in ZO-1cKO. Electrophysiological and optical mapping studies indicated high-grade atrioventricular (A-V) block in ZO-1cKO comparing to CTL hearts. While ZO-1-associated proteins such as vinculin, connexin 43, N-cadherin, and α-catenin showed no significant change with the loss of ZO-1, Connexin-45 and Coxsackie-adenovirus (CAR) proteins were reduced in atria of ZO-1cKO. Further, with loss of ZO-1, ZO-2 protein was increased significantly in ventricular CM in a presumed compensatory manner but was still not detected in the AV nodal myocytes. Importantly, the expression of the sodium channel protein NaV1.5 was altered in AV nodal cells of the ZO-1cKO versus CTL. CONCLUSIONS: ZO-1 protein has a unique physiological role in cardiac nodal tissue. This is in alignment with its known interaction with CAR and Cx45, and a new function in regulating the expression of NaV1.5 in AV node. Uniquely, ZO-1 is dispensable for function of the working myocardium.

RBFox2-miR-34a-Jph2 axis contributes to cardiac decompensation during heart failure.

Heart performance relies on highly coordinated excitation-contraction (EC) coupling, and defects in this critical process may be exacerbated by additional genetic defects and/or environmental insults to cause eventual heart failure. Here we report a regulatory pathway consisting of the RNA binding protein RBFox2, a stress-induced microRNA miR-34a, and the essential EC coupler JPH2. In this pathway, initial cardiac defects diminish RBFox2 expression, which induces transcriptional repression of miR-34a, and elevated miR-34a targets Jph2 to impair EC coupling, which further manifests heart dysfunction, leading to progressive heart failure. The key contribution of miR-34a to this process is further established by administrating its mimic, which is sufficient to induce cardiac defects, and by using its antagomir to alleviate RBFox2 depletion-induced heart dysfunction. These findings elucidate a potential feed-forward mechanism to account for a critical transition to cardiac decompensation and suggest a potential therapeutic avenue against heart failure.

MEMe: A Mutually Enhanced Modeling Method for Efficient and Effective Human Pose Estimation.

In this paper, a mutually enhanced modeling method (MEMe) is presented for human pose estimation, which focuses on enhancing lightweight model performance, but with low complexity. To obtain higher accuracy, a traditional model scale is largely expanded with heavy deployment difficulties. However, for a more lightweight model, there is a large performance gap compared to the former; thus, an urgent need for a way to fill it. Therefore, we propose a MEMe to reconstruct a lightweight baseline model, EffBase transferred intuitively from EfficientDet, into the efficient and effective pose (EEffPose) net, which contains three mutually enhanced modules: the Enhanced EffNet (EEffNet) backbone, the total fusion neck (TFNeck), and the final attention head (FAHead). Extensive experiments on COCO and MPII benchmarks show that our MEMe-based models reach state-of-the-art performances, with limited parameters. Specifically, in the same conditions, our EEffPose-P0 with 256 × 192 can use only 8.98 M parameters to achieve 75.4 AP on the COCO val set, which outperforms HRNet-W48, but with only 14% of its parameters.

Nanocage-Based Capture-Detection System for the Clinical Diagnosis of Autoimmune Disease.

The detection of autoantibodies is critical for diagnosis of autoimmune diseases. However, the sensitivity is often limited by the properties of the antigens and the detection systems such as enzyme-linked immunosorbent assay (ELISA). Here, employing the multidisplay ability of ferritin, a highly sensitive nanocage-based capture-detection system is designed, of which the sensitivity is 100-1000-fold higher than that of conventional ELISA methods. The capture nanocages are constructed by displaying the primary Sjögren's syndrome (pSS)-related antigenic peptides on ferritin nanocage, which present epitopes effectively and high affinity, leading to tenfold higher capture capability for autoantibodies. Human IgG Fc-binding peptides are also engineered on ferritin nanocage, which enable high binding affinity and efficient horseradish peroxidase (HRP)-labeling. Compared with commercial HRP-conjugated anti-human IgG antibody, the nanocage-based detecting probe exhibited more than tenfold increased sensitivity. Autoantibodies are then examined in 91 sera from patients with pSS, 51 from rheumatoid arthritis, 54 from systemic lupus erythematosus, and 55 from healthy individuals by using the nanocage-based ELISA. The results indicate that the nanocage-based capture-detection system is an effective detection platform and provide a novel and more sensitive method for the diagnosis of autoimmune diseases.

Prognostic value of inflammation-based indices in patients with resected hepatocellular carcinoma.

BACKGROUND: As is well recognized that inflammation plays a crucial role in the genesis and progression of various cancer. Here we investigate the prognostic value of a novel index: the combination of neutrophil to lymphocyte ratio and platelet distribution width (coNLR-PDW) in post-operation patients with resectable hepatocellular carcinoma (HCC). METHODS: The receiver operating characteristic (ROC) curve was utilized to determine the optimal cutoff values of continuous variables, including the neutrophil-lymphocyte ratio (NLR) and platelet distribution width (PDW). Kaplan-Meier method and the Log-rank test were used to compare survival differences across three groups stratified by the coNLR-PDW score. Univariate and multivariate Cox proportional hazard regression analyses were adopted to identify independent factors of HCC patient's prognosis. RESULTS: 1.59 and 13.0 were perceived as the optimal cutoff value for NLR and PDW based on the ROC curve, respectively. Kaplan-Meier method revealed that a higher coNLR-PDW score predicts poorer overall survival (OS) and disease-free survival (DFS) (P < 0.001). coNLR-PDW was demonstrated as an independent factor for both OS and DFS using Cox regression analysis in training and validation cohort. CONCLUSION: coNLR-PDW is recognized as a valuable biomarker for predicting the survival of patients with HCC.

BACH1 is transcriptionally inhibited by TET1 in hepatocellular carcinoma in a microRNA-34a-dependent manner to regulate autophagy and inflammation.

Hepatocellular carcinoma (HCC), one of the main contributors to cancer-associated deaths globally, is characterized by high invasiveness. Herein, we studied the molecular mechanisms underlying ten-eleven translocation 1 (TET1)-mediated autophagy in HCC. Following data mining using GSE101728, GSE14520 and GSE138178, TET1 was screened out, and the differential expression of TET1 was verified by bioinformatics analysis. TET1, one of the prognostic markers in HCC, was poorly expressed in HCC. Through functional experiments, we determined that upregulation of TET1 inhibited the proliferation, migration, invasion, tumorigenesis, metastasis and inflammatory factors of HCC cells, and promoted cell autophagy and apoptosis. Mechanistically, TET1 activated miR-34a by demethylating miR-34a. BTB domain and CNC homology 1 (BACH1) was identified as the target gene of miR-34a. Notably, Downregulation of miR-34a increased cellular inflammatory factors and decreased autophagy in the presence of TET1, while declines in BACH1 suppressed cellular inflammatory factors and enhanced autophagy in the presence of miR-34a inhibitor. BACH1 negatively regulated the p53 pathway. In conclusion, TET1 is a tumor suppressor in the progression of HCC by regulating the miR-34a/BACH1/p53 axis, and may contribute to the improvement of HCC prognosis and therapy.

Shallow Graph Convolutional Network for Skeleton-Based Action Recognition.

Graph convolutional networks (GCNs) have brought considerable improvement to the skeleton-based action recognition task. Existing GCN-based methods usually use the fixed spatial graph size among all the layers. It severely affects the model's abilities to exploit the global and semantic discriminative information due to the limits of receptive fields. Furthermore, the fixed graph size would cause many redundancies in the representation of actions, which is inefficient for the model. The redundancies could also hinder the model from focusing on beneficial features. To address those issues, we proposed a plug-and-play channel adaptive merging module (CAMM) specific for the human skeleton graph, which can merge the vertices from the same part of the skeleton graph adaptively and efficiently. The merge weights are different across the channels, so every channel has its flexibility to integrate the joints. Then, we build a novel shallow graph convolutional network (SGCN) based on the module, which achieves state-of-the-art performance with less computational cost. Experimental results on NTU-RGB+D and Kinetics-Skeleton illustrates the superiority of our methods.

Blind Deblurring Based on Sigmoid Function.

Blind image deblurring, also known as blind image deconvolution, is a long-standing challenge in the field of image processing and low-level vision. To restore a clear version of a severely degraded image, this paper proposes a blind deblurring algorithm based on the sigmoid function, which constructs novel blind deblurring estimators for both the original image and the degradation process by exploring the excellent property of sigmoid function and considering image derivative constraints. Owing to these symmetric and non-linear estimators of low computation complexity, high-quality images can be obtained by the algorithm. The algorithm is also extended to image sequences. The sigmoid function enables the proposed algorithm to achieve state-of-the-art performance in various scenarios, including natural, text, face, and low-illumination images. Furthermore, the method can be extended naturally to non-uniform deblurring. Quantitative and qualitative experimental evaluations indicate that the algorithm can remove the blur effect and improve the image quality of actual and simulated images. Finally, the use of sigmoid function provides a new approach to algorithm performance optimization in the field of image restoration.

Robust Global Motion Estimation for Video Stabilization Based on Improved K-Means Clustering and Superpixel.

Obtaining accurate global motion is a crucial step for video stabilization. This paper proposes a robust and simple method to implement global motion estimation. We don't extend the framework of 2D video stabilization but add a "plug and play" module to motion estimation based on feature points. Firstly, simple linear iterative clustering (SLIC) pre-segmentation is used to obtain superpixels of the video frame, clustering is performed according to the superpixel centroid motion vector and cluster center with large value is eliminated. Secondly, in order to obtain accurate global motion estimation, an improved K-means clustering is proposed. We match the feature points of the remaining superpixels between two adjacent frames, establish a feature points' motion vector space, and use improved K-means clustering for clustering. Finally, the richest cluster is being retained, and the global motion is obtained by homography transformation. Our proposed method has been verified on different types of videos and has efficient performance than traditional approaches. The stabilization video has an average improvement of 0.24 in the structural similarity index than the original video and 0.1 higher than the traditional method.

Robust Template Adjustment Siamese Network for Object Visual Tracking.

Most of the existing trackers address the visual tracking problem by extracting an appearance template from the first frame, which is used to localize the target in the current frame. Unfortunately, they typically face the model degeneration challenge, which easily results in model drift and target loss. To address this issue, a novel Template Adjustment Siamese Network (TA-Siam) is proposed in this paper. The proposed framework TA-Siam consists of two simple subnetworks: The template adjustment subnetwork for feature extraction and the classification-regression subnetwork for bounding box prediction. The template adjustment module adaptively uses the feature of subsequent frames to adjust the current template. It makes the template adapt to the target appearance variation of long-term sequence and effectively overcomes model drift problem of Siamese networks. In order to reduce classification errors, the rhombus labels are proposed in our TA-Siam. For more efficient learning and faster convergence, our proposed tracker uses a more effective regression loss in the training process. Extensive experiments and comparisons with trackers are conducted on the challenging benchmarks including VOT2016, VOT2018, OTB50, OTB100, GOT-10K, and LaSOT. Our TA-Siam achieves state-of-the-art performance at the speed of 45 FPS.

Filter Pruning via Measuring Feature Map Information.

Neural network pruning, an important method to reduce the computational complexity of deep models, can be well applied to devices with limited resources. However, most current methods focus on some kind of information about the filter itself to prune the network, rarely exploring the relationship between the feature maps and the filters. In this paper, two novel pruning methods are proposed. First, a new pruning method is proposed, which reflects the importance of filters by exploring the information in the feature maps. Based on the premise that the more information there is, more important the feature map is, the information entropy of feature maps is used to measure information, which is used to evaluate the importance of each filter in the current layer. Further, normalization is used to realize cross layer comparison. As a result, based on the method mentioned above, the network structure is efficiently pruned while its performance is well reserved. Second, we proposed a parallel pruning method using the combination of our pruning method above and slimming pruning method which has better results in terms of computational cost. Our methods perform better in terms of accuracy, parameters, and FLOPs compared to most advanced methods. On ImageNet, it is achieved 72.02% top1 accuracy for ResNet50 with merely 11.41M parameters and 1.12B FLOPs.For DenseNet40, it is obtained 94.04% accuracy with only 0.38M parameters and 110.72M FLOPs on CIFAR10, and our parallel pruning method makes the parameters and FLOPs are just 0.37M and 100.12M, respectively, with little loss of accuracy.