Multiple Branches Research Articles

Improving the Detection and Positioning of Camouflaged Objects in YOLOv8

Camouflaged objects can be perfectly hidden in the surrounding environment by designing their texture and color. Existing object detection models have high false-negative rates and inaccurate localization for camouflaged objects. To resolve this, we improved the YOLOv8 algorithm based on feature enhancement. In the feature extraction stage, an edge enhancement module was built to enhance the edge feature. In the feature fusion stage, multiple asymmetric convolution branches were introduced to obtain larger receptive fields and achieve multi-scale feature fusion. In the post-processing stage, the existing non-maximum suppression algorithm was improved to address the issue of missed detection caused by overlapping boxes. Additionally, a shape-enhanced data augmentation method was designed to enhance the model’s shape perception of camouflaged objects. Experimental evaluations were carried out on camouflaged object datasets, including COD and CAMO, which are publicly accessible. The improved method exhibits enhancements in detection performance by 8.3% and 9.1%, respectively, compared to the YOLOv8 model.

Cryptotermespugnus (Blattodea, Isoptera, Kalotermitidae), a new drywood termite species from the Brazilian Caatinga dry forest and key to South American Cryptotermes Banks, 1909.

A new termite species, Cryptotermespugnussp. nov., is described from northeastern Brazil. The winged imago of C.pugnus is distinguished from most congeners by the lack of arolia and the multiple branches connecting the median vein to the radial sector. The soldier is unique among South American Cryptotermes by its cuboidal head capsule and very rugose postclypeus. The new species constitutes the fourteenth Cryptotermes species on the continent for which we provide a key to soldiers.

Simultaneous observations of equatorial plasma bubbles with an all-sky airglow imager and a HF Doppler sounding system in Taiwan

High-Frequency Doppler (HFD) sounders at low-latitudes often detect characteristic oblique spreading Doppler traces in the spectrogram, known as Oblique Spread Structure (OSS). OSS has been expected to be generated by the dispersion of radio wave reflection due to equatorial plasma bubbles (EPBs). However, it has not yet been confirmed whether OSS is surely a manifestation of EPB by conducting simultaneous observations of EPB and OSS with different observational techniques. Additionally, it remains unclear what kinds of properties of EPB are reflected in the fine structure of OSS. In this study, we investigated three cases of OSSs and EPBs simultaneously observed by a HFD sounding system and an all-sky airglow imager in Taiwan. For the three cases presented here, the timing of OSS occurrence in the HFD data well coincided with that of the EPB appearance in the airglow data. The frequency shift of OSS is quantitatively explained assuming a radio wave reflection at 250–300 km altitudes. These results strongly indicate that OSS is formed by electron density variations at F-region altitudes accompanying EPB; thus, OSS is a manifestation of EPB in the HFD observations. Furthermore, it was suggested that the fine structure of OSS reflected the branching structure of EPB when the multiple branches of EPB reached the intermediate reflection point of the HFD observation. The detection of EPB occurrence and its fine structure using HFD observation enables monitoring of EPB regardless of weather conditions, which will contribute to monitoring the space weather impact of EPBs, for example, on GNSS navigation, in a wide area.Graphical

HS-YOLO: Small Object Detection for Power Operation Scenarios

Object detection methods are commonly employed in power safety monitoring systems to detect violations in surveillance scenes. However, traditional object detection methods are ineffective for small objects that are similar to the background information in the power monitoring scene, which consequently affects the performance of violation behavior detection. This paper proposed a small object detection algorithm named HS-YOLO, based on High-Resolution Network (HRNet) and sub-pixel convolution. First, to fully extract the microfeature information of the object, a small object feature extraction backbone network is proposed based on the HRNet structure. The feature maps of different scales are processed by multiple parallel branches and fused with each other in the network. Then, to fully retain the effective features of small objects, the sub-pixel convolution module is incorporated as the upsampling operator in the feature fusion network. The low-resolution feature map is upsampled to a higher resolution by reorganizing pixel values and performing padding operations in this module. On our self-constructed power operation dataset, the HS-YOLO algorithm achieved a mAP of 87.2%, which is a 3.5% improvement compared to YOLOv5. Particularly, the dataset’s AP for detecting small objects such as cuffs, necklines, and safety belts is improved by 10.7%, 5.8%, and 4.4%, respectively. These results demonstrate the effectiveness of our proposed method in detecting small objects in power operation scenarios.

Improving two-mode algorithm via probabilistic selection for solving satisfiability problem

The satisfiability problem (SAT) is a critically important issue in multiple branches of computer science and artificial intelligence, with its relevance in industrial applications being of particular Significance CCAnr is the current leading stochastic local search (SLS) solver for tackling crafted satisfiable instances. It uses a two-mode strategy, greedy mode and diversification mode. In the present work, we employ a probabilistic selection approach to enhance CCAnr, leading to a new algorithm called ProbCCAnr. Experiments are carried out using the random SAT benchmarks and structured SAT benchmarks including instances encoded from mathematical problems and application problems. The experiments demonstrate that ProbCCAnr significantly improves the performance of state-of-the-art SLS algorithms including CCAnr and ProbSAT, among others. Moreover, ProbCCAnr shows better performance than state of the art complete solvers.

OR20-05 Generation Of CDH23-mutated Pituitary Neuroendocrine Tumor Organoids From Induced Pluripotent Stem Cells To Model Cushing’s Disease Pathogenesis

Abstract Disclosure: J. Chakrabarti: None. S. Mallick: None. J. Churko: None. J. Eschbacher: None. A.S. Little: None. K.C. Yuen: None. Y. Zavros: None. Background: Cushing's disease (CD) is a debilitating disease caused by an adrenocorticotropic hormone (ACTH)-secreting pituitary tumor (PitNET). While most corticotropic PitNETs occur sporadically, rarely CD arises in the context of genetic familial syndromes. CDH23 syndrome is clinically associated with the development of Usher syndrome. In a study that linked mutations in CDH23 with familial and sporadic PitNETs suggested that these genetic alterations could play a role in the pathogenesis of CD. Patient-derived induced pluripotent stem cells (iPSCs) resemble germline mutations of affected individuals, can be differentiated into several tissues, and are used to study hereditary diseases. Our published studies (Mallick et al., 2023, Translational Research) demonstrated that organoids generated from a CD patient expressing CDH23 mutation (iPSCCDH23 organoids) exhibited tumor behavior consistent with the expected pathology and function of corticotroph PitNET from patients. Objective: To generate iPSCs from CD patients expressing CDH23 (iPSCCDH23) to reveal the disease pathogenesis under directed differentiation to PitNET organoids. Methods: Single cell RNA sequencing (scRNA-seq) was used to compare the transcriptomic profile of iPSCCDH23, iPSC organoids generated from a healthy individual (hPitiPSC), and patient PitNET tissue generated organoids (hPitNET41org and hPitNET42org). Lineage trajectory analysis was used on scRNA-seq data collected at 15, 30 and 60 days during the iPSC differentiation program. Results: Markers of dysregulated cell cycle and increased proliferation were identified in iPSCCDH23, hPitNET41org and hPitNET42org organoids. The expression of the E2 factor (E2F) family of transcription factors were upregulated in the PitNET organoids compared to hPitiPSC. Noteworthy, was an increase in the CD44/xCT(SLCA7A11) antioxidant system expression in the iPSCCDH23, hPitNET41org and hPitNET42org cultures. Using lineage trajectory analysis using iPSCCDH23 and hPitiPSC cultures, a base cluster enriched in Wnt/NOTCH+ stem cell markers was selected to direct the lineage trajectories for the other clusters. Trajectories for hPitiPSC cultures revealed multiple branching paths, leading to clusters enriched for markers of pituitary cell lineages (lactotroph, somatotroph, thyrotroph, gonadotroph and corticotroph). The trajectories for iPSCCDH23 organoids exhibited two predominant paths from the base cluster to terminal clusters that were highly enriched with markers representative of corticotroph and NOTCH signaling. Conclusion: PitNET organoids generated from iPSCs of CD patients reveal a dysregulation in the cell lineage differentiation pathways leading to a skewed development of corticotrophs. Understanding the molecular mechanisms of the pathogenesis of CD will identify therapeutic targets for patients. Presentation: Saturday, June 17, 2023

Splicing quality control mediated by DHX15 and its G-patch activator SUGP1

Pre-mRNA splicing is surveilled at different stages by quality control (QC) mechanisms. The leukemia-associated DExH-box family helicase hDHX15/scPrp43 is known to disassemble spliceosomes after splicing. Here, using rapid protein depletion and analysis of nascent and mature RNA to enrich for direct effects, we identify a widespread splicing QC function for DHX15 in human cells, consistent with recent invitro studies. We find that suboptimal introns with weak splice sites, multiple branch points, and cryptic introns are repressed by DHX15, suggesting a general role in promoting splicing fidelity. We identify SUGP1 as a G-patch factor that activates DHX15's splicing QC function. This interaction is dependent on both DHX15's ATPase activity and on SUGP1's U2AF ligand motif (ULM) domain. Together, our results support a model in which DHX15 plays a major role in splicing QC when recruited and activated by SUGP1.

Contact of the Tumor with the Internal Mammary Perforator Vessels as Risk Factor for Internal Mammary Lymph Node Involvement Based on Dynamic Contrast Enhancing MRI Imaging in Breast Cancer Patients

Distant metastasis free and overall survival of high-risk breast cancer patients can be improved with irradiation of the lymphatic system. However, inclusion of internal mammary region (IMN) markedly increases the dose to the heart. Hence, an estimation of the individual risk of subclinical involvement is necessary in order to balance the oncological benefit versus the cardiac toxicity. In randomized studies the risk stratification for IMN involvement is mainly based on the tumor location (medial, central vs. lateral). The aim of this study was to evaluate whether the contact of the tumor with the internal mammary perforator vessels (IMPV) is an additional risk factor for IMN involvement. For this trial we retrospectively examined 104 sets of pre-treatment dynamic contrast enhanced magnetic resonance imaging (DE-MRI) scans of locally advanced breast cancer patients. All patients had ipsilateral clinical suspect lymph nodes in the axillary region and either positive (n = 52) or negative lymph nodes (n = 52) in the internal mammary region. Based on the DE-MRI we assessed the tumor location as well as contact to the internal mammary perforating vessels (IMPV). Contact between tumor and IMPV was classified into either predominant, minor or no contact. Criteria for a predominant IMPV contact was defined as: main vessel contact or contact to multiple (> = 5) side branches. The tumor location of IMN+ patients were medial in 19.2% (10/52), central in 26.9% (14/52), lateral in 28.8% (15/52) and multicentric in 25.0% (13/52) of. In IMN- patients, tumor location was medial in 9.6% (5/52), central in 26.9% (14/52), lateral in 42.3% (22/52) and multicentric in 21.2% (11/52) in IMN- patients. The tumor size ranged from T1-T4 with most patients staged as T2 (48.1% (IMN+) vs. 38.5% IMN-). Most patients had hormone receptor positive / Her2neu negative disease (36.5% (IMN+) vs. 46.2% (IMN-). 100 % of tumors in IMN+ patients had contact with IMPV: in 94.2% (49/52) criteria for predominant IMPV contact was fulfilled; in the remaining 3 cases (5.8%) the tumor had minor IMPV contact. In IMN- patients however, 36/52 of cases (69.2%) had direct IMPV contact. Predominant contact was found in only 23.1% (12/52), minor IMPV contact in 46.2% (24/52). Predominant contact of the tumor with IMPV appears to be a potential risk factor for IMN involvement in breast cancer patients and could be used for stratification for IMN-irradiation in future trials.

Dimensional profiling of psychopathology in children and adolescents based on the K-SADS-PL and an analysis of the construct validity of two ADHD symptom dimensions

Objectives The traditional view on psychiatric disorders as categorical and distinct is being challenged by perspectives emphasizing the relevance of dimensional and transdiagnostic assessment. However, most diagnostic instruments are based on a categorical view with a threshold-approach to disease classification. Methods We here describe algorithms for dimensionalizing the psychopathological ratings of the widely used diagnostic interview for children and adolescents, the Kiddie-Schedule for Affective Disorders and Schizophrenia – Present and Lifetime Version (K-SADS-PL). We further evaluate the criterion-related construct validity of the dimensionalized attention-deficit/hyperactivity disorder (ADHD) scales using Rasch models in a sample of 590 children (mean age 10.29 (.36), 49% girls). Results The algorithms generate scores of current symptom load, i.e., the sum of clinician-rated symptoms within each disorder assessed with the interview. We found support for counting symptoms of inattention and hyperactivity/impulsivity, respectively, but not for a single combined ADHD scale. Conclusions The algorithms constitute an initial step in creating a framework for clinician-rated dimensional analyses of symptoms derived from the K-SADS-PL, but future studies are needed to further evaluate the construct validity of the remaining scales and the reliability and clinical utility of the method. We believe that our proposed algorithms offer a novel method of dimensional psychopathological assessment, which can be applied in multiple branches of child and adolescent psychiatry.

Spinning black holes with axion hair

In this work we construct and analyze non-perturbative stationary and axially-symmetric black hole solutions in general relativity coupled to an electromagnetic and an axion field. The axion field is coupled to the electromagnetic field, which leads to hairy solutions in the presence of an electric charge and rotation. We investigate the existence and characteristics of these solutions for different values of the spin, charge and coupling constant. Our analysis shows the presence of violations of the Kerr–Newman bound, solutions with large positive and negative values of the gyromagnetic ratio, and the existence of multiple branches of solutions with distinct properties, demonstrating that black hole uniqueness does not hold in this scenario. The code used in this study is publicly available, providing a valuable tool for further research on this model.

The role of mantle viscosity heterogeneities on the development of secondary plumes in the upper mantle

SUMMARY Recent seismic observations have revealed diverse plume morphologies beneath hotspots that are characterized by multiple secondary plume branches with different sizes and temperatures. However, how the structures of these secondary plumes have developed and what controls the properties of these secondary plumes have not been well explored. Here, we performed 3-D numerical models to systematically investigate the effects of mantle viscosity structures and the strength of the 660-km phase change on the development of secondary plumes. We find that the characteristics of the secondary plumes, in terms of their numbers, excess temperatures and sizes, are controlled by the Clapeyron slope of the 660-km phase change (γ660), the properties of the low-viscosity layer beneath the mantle transition zone (viscosity reduction ratio, thickness, depth or phase change dependency of viscosity) and the presence of the weak asthenosphere. For γ660 between −2.0 and −3.0 MPa K−1, the weak layer viscosity reduction ratio plays a first-order role in controlling secondary plume properties, that is, a smaller viscosity reduction ratio results in more secondary plumes, higher plume excess temperature and larger plume size. The thickness and viscosity dependencies of the weak layer have a secondary effect on secondary plume properties. However, a relatively small γ660 (e.g. −1.0 MPa K−1) would result in a single secondary plume irrespective of the weak layer viscosity reduction ratio, the viscosity dependency, and whether with or without the weak asthenosphere. However, our models also show that the presence of the weak asthenosphere has a prominent influence in increasing the number of secondary plumes but reducing the size of the plumes. Our results provide new insight into plume dynamics and suggest that the diverse plume structures observed in the upper mantle likely reflect different mantle viscosity structures beneath the hotspots.

Numerical accuracy of principal geodesic analysis on the sphere in director‐based dynamics of hybrid mechanical systems

AbstractWe aim at nonlinear model order reduction (MOR) in hybrid mechanical systems by means of Principal Geodesic Analysis (PGA) on the Riemannian manifolds S2 (the sphere) and SO(3) (the rotation group). MOR requires highly accurate and efficient implementations of the logarithm maps and the resulting lifts across multiple branches. However, in our cases these maps have singularities due to periodicity. In this work we focus on the logarithm and lift maps for the sphere S2. We conduct detailed numerical experiments on mechanical systems to achieve maximal accuracy in spite of the singularities.

Multi-Aperture Coherent Digital Combining Based on Complex-Valued MIMO 2N × 2 Adaptive Equalizer for FSO Communication

Collecting light using multiple small apertures offers a lot of advantages over doing so using a single large aperture, including the effective mitigation of scintillation under strong turbulence, low implementation cost, and ease of scalability. However, the tricky of a multi-aperture system is to combine the multiple branches signals with static skew mismatch and with time-varying characteristics of gain, phase and state of polarization, especially for coherent digital combining of polarization multiplexing (PM) signals. In this paper, a complex-valued multiple-input multiple-output (MIMO) 2N×2 adaptive equalizer is designed to combine multi-aperture signals, and the tap coefficient of each branch is calculated based on the constant modulus algorithm (CMA). This technique relaxes hardware constraints on the time alignment of multi-aperture branch since the skew can be compensated electronically thanks to the proposed blind digital combining algorithm. In addition, laser linewidth within evaluation range smaller than 10 MHz has no effect on the performance of 2N×2 adaptive equalizer. Moreover, the feasibility of the proposed adaptive equalizer for moderate and strong turbulence suppression is verified by a 10-Gbps data rate PM-QPSK modulation offline two- and four-aperture simulated turbulence experiment.

DeepAFP: An effective computational framework for identifying antifungal peptides based on deep learning.

Fungal infections have become a significant global health issue, affecting millions worldwide. Antifungal peptides (AFPs) have emerged as a promising alternative to conventional antifungal drugs due to their low toxicity and low propensity for inducing resistance. In this study, we developed a deep learning-based framework called DeepAFP to efficiently identify AFPs. DeepAFP fully leverages and mines composition information, evolutionary information, and physicochemical properties of peptides by employing combined kernels from multiple branches of convolutional neural network with bi-directional long short-term memory layers. In addition, DeepAFP integrates a transfer learning strategy to obtain efficient representations of peptides for improving model performance. DeepAFP demonstrates strong predictive ability on carefully curated datasets, yielding an accuracy of 93.29% and an F1-score of 93.45% on the DeepAFP-Main dataset. The experimental results show that DeepAFP outperforms existing AFP prediction tools, achieving state-of-the-art performance. Finally, we provide a downloadable AFP prediction tool to meet the demands of large-scale prediction and facilitate the usage of our framework by the public or other researchers. Our framework can accurately identify AFPs in a short time without requiring significant human and material resources, and hence can accelerate the development of AFPs as well as contribute to the treatment of fungal infections. Furthermore, our method can provide new perspectives for other biological sequence analysis tasks.

A structurally re-parameterized convolution neural network-based method for gearbox fault diagnosis in edge computing scenarios

Gearboxes operate in harsh environments. Cloud-based techniques have been previously adopted for fault diagnosis in Gearboxes. Cloud-based fault diagnosis methods are prone to time delays and loss of information. Therefore, edge computing-based fault diagnosis becomes an option. However, with limited hardware resources for edge devices, balancing the diagnostic capabilities of the model with operating performance becomes a challenge. This paper proposes a lightweight convolutional neural network for gearbox fault diagnosis in edge computing scenarios to achieve an accurate diagnosis and lightweight deployment of models. By constructing the Mel-Frequency Cepstral Coefficients (MFCC) feature matrix of input data, the methodology can suppress noise interference and improve diagnostic accuracy. Based on the structural re-parameterization, the model structure transforms from multiple branches at training time to a single branch at inference time. This improves the inference speed of the model and reduces the hardware cost when the model is deployed while ensuring that the diagnostic capability of the model remains unchanged. Validation experiments were conducted on a public dataset and a custom experimental device, using the NVIDIA Jetson Xavier NX kit as the edge computing platform. According to the experiment result, after extracting the MFCC feature matrix, the average diagnostic accuracy rate in the noisy environment of the presented methodology is improved by 12.22% and 9.44%, respectively. After structural re-parameterization, the Memory of the model decreases by 52.58%, and the inference speed is increased by 38.83%.

Deep Decoupling Classification and Regression for Visual Tracking

Classification and regression are two tasks that most Siamese-based trackers need to handle. However, most of the existing trackers only learn one feature embedding to handle these two types of task, making it difficult to optimize both simultaneously. To solve this problem, this paper tries to deeply decouple classification and regression in the model structure. Specifically, two feature extraction backbone networks are used to divide the model into two branches to extract the heterogeneous features suitable for the two tasks, respectively. Inspired by the core idea of Transformer, information interaction and fusion between multiple branches are achieved by cross-attention mechanism, which can fully exploit the deep information dependence between multiple branches. In addition, the concept of channel-level information interaction is proposed by innovatively changing the generation mode of vector groups in the attention module. The experiments show that Double Siamese Tracker (DST) designed in this paper greatly improves the accuracy of classification and regression. DST runs at 60 FPS (Frames Per Second) on GPU, far above the real-time requirement.

Channel Estimator and Nonlinear Detector for mmWave Beamformed OTFS Systems in High Mobility Scenarios

Real-time data-intensive applications in highly mobile environments with high data rates and low latency are one of the key requirements for beyond fifth-generation (5 G) communications. Towards this end, a combination of the millimeter wave (mmWave) band with large available spectrum and orthogonal time-frequency space (OTFS) modulation, which offers robust performance in high mobility scenarios, can be envisioned as a promising solution. However, employing OTFS in the mmWave frequency band is challenging in presence of nonlinear distortions attributed to radio frequency (RF) power amplifier (PA) and complex equalization involved in OTFS receiver in the delay-Doppler domain. The nonlinearity effect of PA is inevitable in mmWave systems due to the high operating frequency and large bandwidth. The nonlinear distortions induce interference and make the target distribution analytically intractable. In this paper, we propose a receiver design for mmWave OTFS systems that aims to address the challenges posed by the nonlinear distortions of the RF PA. The proposed receiver design leverages the use of a large-scale antenna array to decouple the received signal into multiple branches through beamforming. Specifically, we first derive the input-output relation in the delay-Doppler domain with nonlinearity. Then, a combination of maximal-ratio combining (MRC) and iterative particle filter (PF)-based nonlinear detector is proposed to address the issue. We also design a pilot pattern considering the additional interference from the PA impairment. Furthermore, we reformulate the threshold-based channel estimation to incorporate the nonlinearly distorted pilot in its estimation process. Extensive numerical results demonstrate the proposed algorithm's reliability and superiority over existing digital predistorters (DPD) and conventional linear detectors.

Numerical simulation of proppant migration and sedimentation behavior in complex fractures based on computational fluid dynamics

The migration and sedimentation behavior of proppant in complex fractures plays a vital role in the formation of fractures with high conductivity. Most of the existing studies on proppant sedimentation and migration are based on a single vertical fracture model that propagates in the same direction. However, due to in situ stress and reservoir heterogeneity, the propagation of fractures in reservoirs will produce multiple branch fractures that intersect with the main fractures and have a certain inclination angle. Therefore, it is necessary to study the sedimentation and migration behavior of proppants in inclined complex fractures. In this paper, the proppant particles sedimentation law in fractures is studied based on computational fluid dynamics. By changing the particle size, volume fraction, and inlet flow velocity of proppant, the migration and sedimentation law of proppant in a single fracture is simulated and verified. Then, through the establishment of a complex fracture model with multiple branches, and from the four dimensions of the vacant length of the front edge of the sand embankment, the length of the sand embankment, the highest point of the sand embankment, and the effective laying area of the proppant, the angle between the main and branch fractures in the complex fractures, the fracture angle and the inlet flow velocity, as well as the location relationship of the branch fracture setting and the migration and sedimentation of the proppant in the complex fractures are explored.

PSLT: A Light-Weight Vision Transformer With Ladder Self-Attention and Progressive Shift.

Vision Transformer (ViT) has shown great potential for various visual tasks due to its ability to model long-range dependency. However, ViT requires a large amount of computing resource to compute the global self-attention. In this work, we propose a ladder self-attention block with multiple branches and a progressive shift mechanism to develop a light-weight transformer backbone that requires less computing resources (e.g., a relatively small number of parameters and FLOPs), termed Progressive Shift Ladder Transformer (PSLT). First, the ladder self-attention block reduces the computational cost by modelling local self-attention in each branch. In the meanwhile, the progressive shift mechanism is proposed to enlarge the receptive field in the ladder self-attention block by modelling diverse local self-attention for each branch and interacting among these branches. Second, the input feature of the ladder self-attention block is split equally along the channel dimension for each branch, which considerably reduces the computational cost in the ladder self-attention block (with nearly [Formula: see text] the amount of parameters and FLOPs), and the outputs of these branches are then collaborated by a pixel-adaptive fusion. Therefore, the ladder self-attention block with a relatively small number of parameters and FLOPs is capable of modelling long-range interactions. Based on the ladder self-attention block, PSLT performs well on several vision tasks, including image classification, objection detection and person re-identification. On the ImageNet-1 k dataset, PSLT achieves a top-1 accuracy of 79.9% with 9.2 M parameters and 1.9 G FLOPs, which is comparable to several existing models with more than 20 M parameters and 4 G FLOPs. Code is available at https://isee-ai.cn/wugaojie/PSLT.html.

Spectral Segmentation Multi-Scale Feature Extraction Residual Networks for Hyperspectral Image Classification

Hyperspectral image (HSI) classification is a vital task in hyperspectral image processing and applications. Convolutional neural networks (CNN) are becoming an effective approach for categorizing hyperspectral remote sensing images as deep learning technology advances. However, traditional CNN usually uses a fixed kernel size, which limits the model’s capacity to acquire new features and affects the classification accuracy. Based on this, we developed a spectral segmentation-based multi-scale spatial feature extraction residual network (MFERN) for hyperspectral image classification. MFERN divides the input data into many non-overlapping sub-bands by spectral bands, extracts features in parallel using the multi-scale spatial feature extraction module MSFE, and adds global branches on top of this to obtain global information of the full spectral band of the image. Finally, the extracted features are fused and sent into the classifier. Our MSFE module has multiple branches with increasing ranges of the receptive field (RF), enabling multi-scale spatial information extraction at both fine- and coarse-grained levels. On the Indian Pines (IP), Salinas (SA), and Pavia University (PU) HSI datasets, we conducted extensive experiments. The experimental results show that our model has the best performance and robustness, and our proposed MFERN significantly outperforms other models in terms of classification accuracy, even with a small amount of training data.