Independent Branches Research Articles

ACP-Net: Asymmetric Center Positioning Network for Real-Time Text Detection

Scene text detection is crucial across numerous application fields. However, despite the emphasis on real-time performance in scene text detection, most existing detection models utilize the Feature Pyramid Network (FPN) for feature extraction, often disregarding its inherent limitations. Integrating high-resolution multi-channel features into FPN requires substantial computational resources. While FPN treats local and global features equally and is stable in various applications, its suitability for text-specific features is questionable. To this end, we propose the Asymmetric Center Positioning Network (ACP-Net) to replace FPN, achieving accuracy and real-time text detection in complex scenarios. ACP-Net features an asymmetric feature structure with independent branches for global and local information, along with an adaptive weighted fusion module to capture long-range dependencies effectively. In addition, a text center positioning module enhances text feature understanding by learning feature centers. Comprehensive evaluations across various terminals confirmed ACP-Net’s superior accuracy and speed.

MBQuant: A novel multi-branch topology method for arbitrary bit-width network quantization

Arbitrary bit-width network quantization has received significant attention due to its high adaptability to various bit-width requirements during runtime. However, in this paper, we investigate existing methods and observe a significant accumulation of quantization errors caused by switching weight and activations bit-widths, leading to limited performance. To address this issue, we propose MBQuant, a novel method that utilizes a multi-branch topology for arbitrary bit-width quantization. MBQuant duplicates the network body into multiple independent branches, where the weights of each branch are quantized to a fixed 2-bit and the activations remain in the input bit-width. For completing the computation of a desired bit-width, MBQuant selects multiple branches, ensuring that the computational costs match those of the desired bit-width, to carry out forward propagation. By fixing the weight bit-width, MBQuant substantially reduces quantization errors caused by switching weight bit-widths. Additionally, we observe that the first branch suffers from quantization errors caused by all bit-widths, leading to performance degradation. Thus, we introduce an amortization branch selection strategy that amortizes the errors. Specifically, the first branch is selected only for certain bit-widths, rather than universally, thereby the errors are distributed among the branches more evenly. Finally, we adopt an in-place distillation strategy that uses the largest bit-width to guide the other bit-widths to further enhance MBQuant’s performance. Extensive experiments demonstrate that MBQuant achieves significant performance gains compared to existing arbitrary bit-width quantization methods. Code is made publicly available at https://github.com/zysxmu/MBQuant.

MSS-JDA: Multi-source self-selected joint domain adaptation method based on cross-subject EEG emotion recognition

The electroencephalogram (EEG) is widely used to identify different emotional states. However, cross-subject emotion recognition based on EEG data is challenging due to physiological differences among participants. Many studies have proposed using multi-source domain adaptation methods to address cross-subject issues and have achieved promising results. multi-source domain adaptation involves constructing independent branches for EEG data from multiple subjects, using one-to-one domain adaptation, and extracting domain-specific features, followed by reasoning through multiple branches. However, a current practical obstacle is the increasing computational time cost as the number of source domains grows. Moreover, too many source domains may lead to domain shift issues. To address these challenges, we propose a novel emotion recognition method based on Multi-Source Domain Branch Self-Selected Joint Domain Adaptation (MSS-JDA). Initially, we construct a shared feature extractor for multiple source domain branches to extract common low-level features. Next, we use a one-to-one associated joint adaptation method to extract features specific to each domain. During training, our method of multi-source domain branch self-selection, guided by prior knowledge in the early training phase, prunes source domain branches that deviate significantly from the target domain, retaining only a certain number of source domain branches close to the target domain for continued training. Finally, domain-specific features are input into their respective classifiers to obtain N predictions, and their average is taken as the final result. Extensive experiments on the SEED and SEED-IV datasets demonstrate that the MSS-JDA model achieves accuracy and standard deviation scores of 93.78%, 3.39 and 78.93%, 7.39 respectively, surpassing most existing models. The experiments indicate that the proposed MSS-JDA method improves the accuracy of EEG emotion recognition classification, thus benefiting practical applications in EEG classification.

Cross pathogenicity, host range and molecular characteristics of Fusarium oxysporum species complex populations isolated from tobacco in Jilin Province, China

AbstractFusarium wilt of tobacco (FWT) is a globally significant disease responsible for major losses in tobacco yield and quality. A total of 320 Fusarium oxysporum species complex (FOSC) isolates derived from five tobacco‐growing regions in Jilin province, China, have been confirmed as the predominant pathogens causing FWT, including F. tobaccum, F. cugenangense, F. odoratissimum and F. foetens. In this study, the results indicated that differences in the virulence of the representative isolates from tobacco from the same Fusarium species were significant. Aside from infecting tobacco, the virulence of these representative FOSC isolates differed on chilli and other crops, and they were not pathogenic or were weakly pathogenic to a further 20 crops. Six strains of formae speciales (ff. sp.) of F. oxysporum had the strongest virulence on their original hosts and were not or weakly pathogenic to tobacco. Joint analysis of the pg1‐pg5‐pgx1‐pgx4 genes indicated that 21 isolates of F. cugenangense were divided into four branches, which were clustered with different ff. sp. of F. oxysporum; five isolates of F. foetens were clustered in a branch with f. sp. fragariae, which were not pathogenic to strawberries; six isolates of F. tobaccum and five isolates of F. odoratissimum comprised independent branches and did not cluster with any other f. sp. of F. oxysporum. In sum, none of the FOSC isolates causing FWT in Jilin exhibited tobacco specificity. Our findings further confirmed that FOSC isolates from tobacco could be distinguished into F. tobaccum, F. cugenangense, F. odoratissimum and F. foetens.

Research on angular displacement detection technology of inter-satellite laser communication based on coherent system

With the rapid development of spatial information networks technologies, inter-satellite laser communication links have played an important role in the satellite Internet. High-precision relative angular displacement detection is the foundation for achieving real-time tracking of laser communication links, and its accuracy directly impacts the tracking performance and communication quality of the laser communication link. Traditional laser communication systems utilize intensity-based position detection methods to calculate angular displacement, requiring setting separate Charge Coupled Device (CCD) or Four-quadrant Detector (QD) as well as independent optical tracking branches. This setup makes it difficult to integrate with high-speed coherent communication systems. Additionally, the method employing intensity detection to determine angular displacement based on spot position has lower accuracy in theory. A angular displacement detection of coherent-based inter-satellite laser communication is proposed in this paper. Firstly, real-time calculation of phase differences at different positions of the beam is achieved through heterodyne coherent array detection and dual high-precision phase-locked loops. Then, a model correlating wave front phase difference with relative angular displacement is established, phase noise during beam transmission and detection processes is analyzed, and algorithm for detecting angular displacement in coherent systems is studied. Finally, a laser communication link is constructed under laboratory condition, and experimental validation work is carried out. When the relative angle deviation is 1°, the phase difference changes by 0.089049795 rad, and the angular displacement detection error of 0.66697μrad(σ). When the relative angle error is 0.603873μrad(σ), the detection sensitivity is −43.93dBm. The experimental results show that in the coherent system, the relative angular displacement detection with high precision can be realized by using the wavefront phase difference of beam, which can support the laser communication tracking system to achieve higher precision beam tracking.

Cross-Modal Face Super-Resolution Based on Quasi-Siamese Domain Transfer Fusion Network

In this paper, we propose a Cross-Modal Face Super-Resolution (CMFSR) method to construct high-resolution (HR) facial images from low-resolution (LR) cross-modal facial images captured respectively by disjoint visible light (VIS) and near-infrared (NIR) cameras. Due to the coupling of modality transformation and information fusion, CMFSR is more difficult to obtain HR reconstructed results compared with traditional super-resolution. To solve this problem, a Quasi-Siamese Domain Transfer Fusion Network (QSDTFN) for CMFSR is proposed in this paper, whose two branches transfer two LR face modality to HR face modality by domain transfer respectively. Different from two completely independent branches in the traditional pseudo-siamese network, only the HR-to-LR face transfer processes of the two branches in our quasi-siamese network are independent, while the LR-to-HR face transfer processes are coupled. This coupled module called the Adaptive Weighted Domain Transfer Fusion Module (AWDTFM) disentangles the modality and identity information in the two LR faces, thus achieving modality transformation and identity information fusion simultaneously. In order to strengthen the optimization on the process of CMFSR, this method further introduces the backward QSDTFN to form a higher-level bidirectional structure with the forward QSDTFN, and specifically designs two types of losses: intra-network loss and inter-network loss, to constrain the modality and identity consistencies within one QSDTFN and between two QSDTFNs respectively. The experimental results on the challenging LR cross-modal face datasets demonstrate that the proposed method performs favorably against the state-of-the-art methods.

Magnetic phase transition and critical behavior in (Pr,Sm)[formula omitted]Fe[formula omitted] compounds: A comprehensive study using macroscopic and local experimental approaches

Achieving a comprehensive understanding of the properties and behaviors of magnetic phase transitions and critical behavior is a key objective of this study. A macroscopic approach utilizing magnetization measurement and a local approach utilizing 57Fe Mössbauer spectrometry were employed. The findings indicate that the critical exponents of Pr2Fe17 and (Pr,Sm)2Fe17 closely approximate those of the 3D-Ising model, as confirmed by the scaling analysis of our critical exponents. Moreover, we observed that isotherms collapse into two distinct and independent universal branches above and below the Curie temperature when utilizing the critical exponents we obtained. Our results indicate that the exponents determined using both conventional methods and Mössbauer spectrometry closely match those of the 3D-Ising model for spins coupled in three dimensions (lattice dimensionality d=3, spin dimensionality n=1) with attractive interactions between spins (J(r)) at the boundary of short-range and long-range decay as J(r)∝r−(3+σ).

Loss Minimization Model Predictive Current Control for Linear Induction Motors

To further improve the operating efficiency of linear induction motor, this paper proposes a loss minimization model predictive current control method based on governable loss online calculation. First, the method derives the equivalent circuit of linear induction motor containing independent iron loss branches, and establishes the loss model containing iron loss. Second, the expressions of stator d‐axis current with governable losses and secondary flux amplitude are derived from theoretical analysis, and then a given value of stator d‐axis current is derived when the governable losses are minimal. Then, a chain observer is designed to observe the secondary flux and calculate the governable loss and stator d‐axis current under the current operation. Finally, a method is proposed to introduce the stator d‐axis current observation results into the model predictive current control, which ensures the loss suppression effect and improves the dynamic performance at the same time. In addition, this article also compares the loss minimization controllers with or without iron loss. The simulation results are consistent with the theory, and the optimal control of the secondary flux of the linear induction motor can be successfully realized, thus effectively reducing the energy loss of the motor in the dynamic operation process. © 2024 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

Hydrodynamics of Borromean Counterfluids.

Counterflow superfluidity in a system with N≥3 components is distinctively different from the N=2 case. The key feature is the difference between the number (N) of elementary vortex excitations and the number (N-1) of independent branches of phonon modes, that is, the number of superfluid modes is larger than the number of ordered phase variables. We formulate a hydrodynamic theory of this state. We show how all the dynamical and statistical aspects of this ("Borromean") type of ordering are naturally described by effective N-component theory featuring compact-gauge invariance. We also discuss how off diagonal intercomponent couplings convert the Borromean supercounterfluid into a Borromean insulator, with an emphasis on the properties of a nontrivial state with broken time-reversal symmetry.

Nocardioides bizhenqiangii sp. nov. and Nocardioides renjunii sp. nov., isolated from soil and faeces of Tibetan antelope (Pantholops hodgsonii) on the Qinghai-Tibet Plateau

Two novel strain pairs (HM61T/HM23 and S-34T/S-58) were isolated from soil and the faeces of Tibetan antelope (Pantholops hodgsonii) collected at the Qinghai-Tibet Plateau of PR China. All four new isolates were aerobic, non-motile, Gram-stain-positive, catalase-positive, oxidase-negative, and short rod-shaped bacteria. The results of phylogenetic analysis based on the full-length 16S rRNA genes and 283 core genomic genes indicated that the four strains were separated into two independent branches belonging to the genus Nocardioides. Strains HM61T and HM23 were most closely related to Nocardioides pelophilus THG T63T (98.58 and 98.65 % 16S rRNA gene sequence similarity). Strains S-34T and S-58 were most closely related to Nocardioides okcheonensis MMS20-HV4-12T (98.89 and 98.89 % 16S rRNA gene sequence similarity). The G+C contents of the genomic DNA of strains HM61T and S-34T were 70.6 and 72.5 mol%, respectively. Strains HM61T, S-34T and the type strains of closely related species in the analysis had average nucleotide identity values of 75.4–90.5 % as well as digital DNA–DNA hybridization values between 20.1 and 40.8 %, which clearly indicated that the four isolates represent two novel species within the genus Nocardioides. The chemotaxonomic characteristics of strains HM61T and S-34T were consistent with the genus Nocardioides. The major fatty acids of all four strains were iso-C16 : 0, C17 : 1 ω8c or C18 : 1 ω9c. For strains HM61T and S-34T, MK-8(H4) was the predominant respiratory quinone, ll-2,6-diaminopimelic acid was the diagnostic diamino acid in the cell-wall peptidoglycan, and the polar lipids profiles were composed of diphosphatidylglycerol and phosphatidylglycerol. Based on phylogenetic, phenotypic, and chemotaxonomic data, we propose that strains HM61T and S-34T represent two novel species of the genus Nocardioides, respectively, with the names Nocardioides bizhenqiangii sp. nov. and Nocardioides renjunii sp. nov. The type strains are HM61T (=GDMCC 4.343T=JCM 36399T) and S-34T (=CGMCC 4.7664T=JCM 33792T).

A Scene Classification Model Based on Global-Local Features and Attention in Lie Group Space

The efficient fusion of global and local multi-scale features is quite important for remote sensing scene classification (RSSC). The scenes in high-resolution remote sensing images (HRRSI) contain many complex backgrounds, intra-class diversity, and inter-class similarities. Many studies have shown that global features and local features are helpful for RSSC. The receptive field of a traditional convolution kernel is small and fixed, and it is difficult to capture global features in the scene. The self-attention mechanism proposed in transformer effectively alleviates the above shortcomings. However, such models lack local inductive bias, and the calculation is complicated due to the large number of parameters. To address these problems, in this study, we propose a classification model of global-local features and attention based on Lie Group space. The model is mainly composed of three independent branches, which can effectively extract multi-scale features of the scene and fuse the above features through a fusion module. Channel attention and spatial attention are designed in the fusion module, which can effectively enhance the crucial features in the crucial regions, to improve the accuracy of scene classification. The advantage of our model is that it extracts richer features, and the global-local features of the scene can be effectively extracted at different scales. Our proposed model has been verified on publicly available and challenging datasets, taking the AID as an example, the classification accuracy reached 97.31%, and the number of parameters is 12.216 M. Compared with other state-of-the-art models, it has certain advantages in terms of classification accuracy and number of parameters.

SwinDPSR: Dual-Path Face Super-Resolution Network Integrating Swin Transformer

Whether to use face priors in the face super-resolution (FSR) methods is a symmetry problem.Various face priors are used to describe the overall and local face features, making the generation of super-resolution face images expensive and laborious. FSR methods that do not require any prior information tend to focus too much on the local features of the face, ignoring the modeling of global information. To solve this problem, we propose a dual-path facial image super-resolution network (SwinDPSR) fused with Swin Transformer. The network does not require additional face priors, and it learns global face shape and local face components through two independent branches. In addition, the channel attention ECA module is used to aggregate the global and local face information in the above dual-path sub-networks, which can generate corresponding high-quality face images. The results of face super-resolution reconstruction experiments on public face datasets and a real-scene face dataset show that SwinDPSR is superior to previous advanced methods both in terms of visual effects and objective indicators. The reconstruction results are evaluated with four evaluation metrics: peak signal-to-noise ratio (PSNR), structural similarity (SSIM), learned perceptual image patch similarity (LPIPS), and mean perceptual score (MPS).

Specificity and dynamics of H2O2 detoxification by the cytosolic redox regulatory network as revealed by in vitro reconstitution

The thiol redox state is a decisive functional characteristic of proteins in cell biology. Plasmatic cell compartments maintain a thiol-based redox regulatory network linked to the glutathione/glutathione disulfide couple (GSH/GSSG) and the NAD(P)H system. The basic network constituents are known and in vivo cell imaging with gene-encoded probes have revealed insight into the dynamics of the [GSH]2/[GSSG] redox potential, cellular H2O2 and NAD(P)H+H+ amounts in dependence on metabolic and environmental cues. Less understood is the contribution and interaction of the network components, also because of compensatory reactions in genetic approaches. Reconstituting the cytosolic network of Arabidopsis thaliana in vitro from fifteen recombinant proteins at in vivo concentrations, namely glutathione peroxidase-like (GPXL), peroxiredoxins (PRX), glutaredoxins (GRX), thioredoxins, NADPH-dependent thioredoxin reductase A and glutathione reductase and applying Grx1-roGFP2 or roGFP2-Orp1 as dynamic sensors, allowed for monitoring the response to a single H2O2 pulse. The major change in thiol oxidation as quantified by mass spectrometry-based proteomics occurred in relevant peptides of GPXL, and to a lesser extent of PRX, while other Cys-containing peptides only showed small changes in their redox state and protection. Titration of ascorbate peroxidase (APX) into the system together with dehydroascorbate reductase lowered the oxidation of the fluorescent sensors in the network but was unable to suppress it. The results demonstrate the power of the network to detoxify H2O2, the partially independent branches of electron flow with significance for specific cell signaling and the importance of APX to modulate the signaling without suppressing it and shifting the burden to glutathione oxidation.

LRFNet: A real-time medical image fusion method guided by detail information

Multimodal medical image fusion (MMIF) technology plays a crucial role in medical diagnosis and treatment by integrating different images to obtain fusion images with comprehensive information. Deep learning-based fusion methods have demonstrated superior performance, but some of them still encounter challenges such as imbalanced retention of color and texture information and low fusion efficiency. To alleviate the above issues, this paper presents a real-time MMIF method, called a lightweight residual fusion network. First, a feature extraction framework with three branches is designed. Two independent branches are used to fully extract brightness and texture information. The fusion branch enables different modal information to be interactively fused at a shallow level, thereby better retaining brightness and texture information. Furthermore, a lightweight residual unit is designed to replace the conventional residual convolution in the model, thereby improving the fusion efficiency and reducing the overall model size by approximately 5 times. Finally, considering that the high-frequency image decomposed by the wavelet transform contains abundant edge and texture information, an adaptive strategy is proposed for assigning weights to the loss function based on the information content in the high-frequency image. This strategy effectively guides the model toward preserving intricate details. The experimental results on MRI and functional images demonstrate that the proposed method exhibits superior fusion performance and efficiency compared to alternative approaches. The code of LRFNet is available at https://github.com/HeDan-11/LRFNet.

Switchable and tunable dual-wavelength thulium-doped fiber laser based on the parallel filter of the all-fiber Sagnac interferometer embedded with an FBG and the Fabry-Perot filter

A switchable and tunable dual-wavelength thulium-doped fiber laser (ST-DW-TDFL) based on the parallel filter is proposed and experimentally demonstrated in this study. In the proposed ST-DW-TDFL, there are two independent branches to form the parallel filter. One branch uses an all-fiber Sagnac interferometer embedded with a fiber Bragg grating (FBG-FSI) as the wavelength selection element, and the other branch uses a Fabry-Perot (FP) filter as the wavelength selection element. By adjusting two optical attenuators to change the light power of two branches, single and dual-wavelength output can be achieved and switched. On the basis of achieving dual-wavelength output, independent and continuous tuning of two wavelengths can be achieved by heating the FBG and adjusting the FP filter. The tuning range of one wavelength is up to 5.84 nm, while the tuning range of the other wavelength is up to 99.86 nm. The maximum wavelength variation and power fluctuation are less than 0.03 nm and 0.5 dB at room temperature. And the experimental results show that the proposed ST-DW-TDFL has flexible wavelength switching and tuning capabilities and good wavelength and power stabilities.

Informal Concentration of Powers in Illiberal Constitutionalism: The Case of Hungary

Classic and new concepts of the separation of powers doctrine focus on the control of central political power. Institutional checks and balances are classic legal instruments of this control, especially the activity of independent state organs. In those countries where the political system is dominated by a populist government, the institutional checks and balances and independent institutions are under constant pressure. In a consolidated political system that has already departed from liberal constitutionalism and where constitutional institutions do function only in a formal sense (illiberal constitutionalism), institutional checks and balances lose their significance. However, in addition to the exercise of powers by independent branches of government, other legal institutions or actions can also have a significant impact on relations between branches of government, even informally. Therefore, the system of checks and balances can be described as a multi-layered structure with formal (classical) and informal elements. The article analyses in detail, how the use of informal constitutional amendments and national referendums—as additional institutions belonging to the complex system of separation of powers—instead of controlling the central political power, has contributed to the informal concentration of government power in Hungary in recent years.

Measuring solid particles in sand-carrying gas flow using multiscale vibration response statistics and deep learning algorithms

A method to quantitate sand particles in turbulent gas flow that combines the multiscale triaxial vibration response and deep learning algorithms is proposed. First, an optimized adaptive wavelet-empirical mode decomposition (EMD) denoising method is proposed based on multifrequency coherent and statistical analysis. Second, complex gas–solid turbulent flow information under multiple gas–particle coupling is characterized based on Hilbert-Huang transform (HHT), Hurst analysis, EMD entropy, etc. In addition, a deep learning algorithm that integrates multiscale flow information to determine sand content that includes two independent branches, a pure deep convolutional neural network (CNN) model driven by microscale triaxial response and a shallow long short-term memory (LSTM) network with regularization driven by mesoscale triaxial response, is proposed. Finally, a quantitative model to characterize sand-carrying turbulent gas flow based on the entropy weight effect of the triaxial vibration response is constructed as follows: Csand=A·p.∑i=xzSiQi. Experimental validation indicates that the proposed deep learning algorithm has recognition and prediction accuracy of 97.8% and 96.97% for sand particle size and the power spectrum, respectively, which are higher than those of the existing intelligent models to characterize sand information. Moreover, the quantitative sand content model based on the multiscale response and deep learning algorithm has a maximum error of only 1.56% under strong turbulence.

Identification and characterization of yellow stripe-like genes in maize suggest their roles in the uptake and transport of zinc and iron

BackgroundYellow Stripe-Like (YSL) proteins are involved in the uptake and transport of metal ions. They play important roles in maintaining the zinc and iron homeostasis in Arabidopsis, rice (Oryza sativa), and barley (Hordeum vulgare). However, proteins in this family have not been fully identified and comprehensively analyzed in maize (Zea mays L.).ResultsIn this study, we identified 19 ZmYSLs in the maize genome and analyzed their structural features. The results of a phylogenetic analysis showed that ZmYSLs are homologous to YSLs of Arabidopsis and rice, and these proteins are divided into four independent branches. Although their exons and introns have structural differences, the motif structure is relatively conserved. Analysis of the cis-regulatory elements in the promoters indicated that ZmYSLs might play a role in response to hypoxia and light. The results of RNA sequencing and quantitative real-time PCR analysis revealed that ZmYSLs are expressed in various tissues and respond differently to zinc and iron deficiency. The subcellular localization of ZmYSLs in the protoplast of maize mesophyll cells showed that they may function in the membrane system.ConclusionsThis study provided important information for the further functional analysis of ZmYSL, especially in the spatio-temporal expression and adaptation to nutrient deficiency stress. Our findings provided important genes resources for the maize biofortification.

Hyperspectral Image Classification Using Spectral–Spatial Double-Branch Attention Mechanism

In recent years, deep learning methods utilizing convolutional neural networks have been extensively employed in hyperspectral image classification (HSI) applications. Nevertheless, while a substantial number of stacked 3D convolutions can indeed achieve high classification accuracy, they also introduce a significant number of parameters to the model, resulting in inefficiency. Furthermore, such intricate models often exhibit limited classification accuracy when confronted with restricted sample data, i.e., small sample problems. Therefore, we propose a spectral–spatial double-branch network (SSDBN) with an attention mechanism for HSI classification. The SSDBN is designed with two independent branches to extract spectral and spatial features, respectively, incorporating multi-scale 2D convolution modules, long short-term memory (LSTM), and an attention mechanism. The flexible use of 2D convolution, instead of 3D convolution, significantly reduces the model’s parameter count, while the effective spectral–spatial double-branch feature extraction method allows SSDBN to perform exceptionally well in handling small sample problems. When tested on 5%, 0.5%, and 5% of the Indian Pines, Pavia University, and Kennedy Space Center datasets, SSDBN achieved classification accuracies of 97.56%, 96.85%, and 98.68%, respectively. Additionally, we conducted a comparison of training and testing times, with results demonstrating the remarkable efficiency of SSDBN.

Three new species of the genus Clavulina (Hydnaceae, Cantharellales) from North China based on morphological and phylogenetic analysis.

Clavulina possesses important ecological and economic value and has attracted extensive attention from mycologists. Macrofungal diversity is high in China, but Clavulina species have not been thoroughly studied. In this study, based on morphological evidence and phylogenetic analyses of the nucleotide sequences of three loci (nrITS, nrLSU, and rpb2), three new species of Clavulina from North China were identified. Morphologically, Clavulinachengdeensis is characterized by its white to dirty white basidiomata with somewhat pale orange tips and somewhat wrinkled hymenium. Clavulinagriseoviolacea is characterized by its gray to dark grayish violet basidiomata, with a sometimes-white stipe base, monopodial or irregularly polychotomous toward branch apices. Clavulinapallida is characterized by its white to pale cream white basidiomata with somewhat orange tips. Phylogenetically, the three new species form three independent branches with high support values in the phylogenetic tree.