Similar Features Research Articles

Stacking-Kcr: A Stacking Model for Predicting the Crotonylation Sites of Lysine by Fusing Serial and Automatic Encoder

Background: Protein lysine crotonylation (Kcr), a newly discovered important posttranslational modification (PTM), is typically localized at the transcription start site and regulates gene expression, which is associated with a variety of pathological conditions such as developmental defects and malignant transformation. Objective: Identifying Kcr sites is advantageous for the discovery of its biological mechanism and the development of new drugs for related diseases. However, traditional experimental methods for identifying Kcr sites are expensive and inefficient, necessitating the development of new computational techniques. Methods: In this work, to accurately identify Kcr sites, we propose a model for ensemble learning called Stacking-Kcr. Firstly, extract features from sequence information, physicochemical properties, and sequence fragment similarity. Then, the two characteristics of sequence information and physicochemical properties are fused using automatic encoder and serial, respectively. Finally, the fused two features and sequence fragment similarity features are then respectively input into the four base classifiers, a meta classifier is constructed using the first level prediction results, and the final forecasting results are obtained. Results: The five-fold cross-validation of this model has achieved an accuracy of 0.828 and an AUC of 0.910. This shows that the Stacking-Kcr method has obvious advantages over traditional machine learning methods. On independent test sets, Stacking-Kcr achieved an accuracy of 84.89% and an AUC of 92.21%, which was higher than 1.7% and 0.8% of other state-of-the-art tools. Additionally, we trained Stacking-Kcr on the phosphorylation site, and the result is superior to the current model. Conclusion: These outcomes are additional evidence that Stacking-Kcr has strong application potential and generalization performance.

Lesion-Decoupling-Based Segmentation With Large-Scale Colon and Esophageal Datasets for Early Cancer Diagnosis.

Lesions of early cancers often show flat, small, and isochromatic characteristics in medical endoscopy images, which are difficult to be captured. By analyzing the differences between the internal and external features of the lesion area, we propose a lesion-decoupling-based segmentation (LDS) network for assisting early cancer diagnosis. We introduce a plug-and-play module called self-sampling similar feature disentangling module (FDM) to obtain accurate lesion boundaries. Then, we propose a feature separation loss (FSL) function to separate pathological features from normal ones. Moreover, since physicians make diagnoses with multimodal data, we propose a multimodal cooperative segmentation network with two different modal images as input: white-light images (WLIs) and narrowband images (NBIs). Our FDM and FSL show a good performance for both single-modal and multimodal segmentations. Extensive experiments on five backbones prove that our FDM and FSL can be easily applied to different backbones for a significant lesion segmentation accuracy improvement, and the maximum increase of mean Intersection over Union (mIoU) is 4.58. For colonoscopy, we can achieve up to mIoU of 91.49 on our Dataset A and 84.41 on the three public datasets. For esophagoscopy, mIoU of 64.32 is best achieved on the WLI dataset and 66.31 on the NBI dataset.

The Patent Landscape of mTOR and PTEN Targets.

PTEN and mTOR signaling have many roles, including antiinflammatory, immunosuppressant and cancer. US patents were retrieved to show the current landscape of the mTOR and PTEN targets. PTEN and mTOR targets were analyzed by patent analysis. The U.S. granted patents from January 2003 to July 2022 were performed and analyzed. The results showed that the mTOR target was more attractive in drug discovery than the PTEN target. Our findings indicated that most large global pharmaceutical companies focused the drug discovery related to the mTOR target. The present study demonstrated that mTOR and PTEN targets showed more applications in biological approaches compared to BRAF and KRAS targets. The chemical structures of the inhibitors of the mTOR target demonstrated some similar features to those of the inhibitors of KRAS targets. At this stage, the PTEN target may not be an ideal target subjected to new drug discovery. The present study was the first one which demonstrated that the group of O=S=O may play a critical role in the chemical structures of mTOR inhibitors. It was the first time to show that a PTEN target may be suitably subjected to new therapeutic discovery efforts related to biological applications. Our findings provide a recent insight into therapeutic development for mTOR and PTEN targets.

Target-distractor similarity predicts visual search efficiency but only for highly similar features.

A major constraining factor for attentional selection is the similarity between targets and distractors. When similarity is low, target items can be identified quickly and efficiently, whereas high similarity can incur large costs on processing speed. Models of visual search contrast a fast, efficient parallel stage with a slow serial processing stage where search times are strongly modulated by the number of distractors in the display. In particular, recent work has argued that the magnitude of search slopes should be inversely proportional to target-distractor similarity. Here, we assessed the relationship between target-distractor similarity and search slopes. In our visual search tasks, participants detected an oddball color target among distractors (Experiments 1 & 2) or discriminated the direction of a triangle in the oddball color (Experiment 3). We systematically varied the similarity between target and distractor colors (along a circular CIELAB color wheel) and the number of distractors in the search array, finding logarithmic search slopes that were inversely proportional to the number of items in the array. Surprisingly, we also found that searches were highly efficient (i.e., near-zero slopes) for targets and distractors that were extremely similar (≤20° in color space). These findings indicate that visual search is systematically influenced by target-distractor similarity across different processing stages. Importantly, we found that search can be highly efficient and entirely unaffected by the number of distractors despite high perceptual similarity, in contrast to the general assumption that high similarity must lead to slow and serial search behavior.

DCNet: A Self-Supervised EEG Classification Framework for Improving Cognitive Computing-Enabled Smart Healthcare.

Cognitive computing endeavors to construct models that emulate brain functions, which can be explored through electroencephalography (EEG). Developing precise and robust EEG classification models is crucial for advancing cognitive computing. Despite the high accuracy of supervised EEG classification models, they are constrained by labor-intensive annotations and poor generalization. Self-supervised models address these issues but encounter difficulties in matching the accuracy of supervised learning. Three challenges persist: 1) capturing temporal dependencies in EEG; 2) adapting loss functions to describe feature similarities in self-supervised models; and 3) addressing the prevalent issue of data imbalance in EEG. This study introduces the DreamCatcher Network (DCNet), a self-supervised EEG classification framework with a two-stage training strategy. The first stage extracts robust representations through contrastive learning, and the second stage transfers the representation encoder to a supervised EEG classification task. DCNet utilizes time-series contrastive learning to autonomously construct representations that comprehensively capture temporal correlations. A novel loss function, SelfDreamCatcherLoss, is proposed to evaluate the similarities between these representations and enhance the performance of DCNet. Additionally, two data augmentation methods are integrated to alleviate class imbalances. Extensive experiments show the superiority of DCNet over the current state-of-the-art models, achieving high accuracy on both the Sleep-EDF and HAR datasets. It holds substantial promise for revolutionizing sleep disorder detection and expediting the development of advanced healthcare systems driven by cognitive computing.

Advances in diffuse glioma assessment: preoperative and postoperative applications of chemical exchange saturation transfer.

Chemical Exchange Saturation Transfer (CEST) is a technique that uses specific off-resonance saturation pulses to pre-saturate targeted substances. This process influences the signal intensity of free water, thereby indirectly providing information about the pre-saturated substance. Among the clinical applications of CEST, Amide Proton Transfer (APT) is currently the most well-established. APT can be utilized for the preoperative grading of gliomas. Tumors with higher APTw signals generally indicate a higher likelihood of malignancy. In predicting preoperative molecular typing, APTw values are typically lower in tumors with favorable molecular phenotypes, such as isocitrate dehydrogenase (IDH) mutations, compared to IDH wild-type tumors. For differential diagnosis, the average APTw values of meningiomas are significantly lower than those of high-grade gliomas. Various APTw measurement indices assist in distinguishing central nervous system lesions with similar imaging features, such as progressive multifocal leukoencephalopathy, central nervous system lymphoma, solitary brain metastases, and glioblastoma. Regarding prognosis, APT effectively differentiates between tumor recurrence and treatment effects, and also possesses predictive capabilities for overall survival (OS) and progression-free survival (PFS).

Molecular and Morphological Identification of Sarocladium Species Causing Sheath Rot of Rice in Thailand and Their Division into Physiological Races.

Sheath rot and dirty panicle are some of the major diseases of rice in Thailand. The diseases are traditionally considered to be caused by the pathogen Sarocladium oryzae and damage and lower both the quantity and quality of rice grain. In this study, 32 fungal isolates collected from the central and northeastern regions of Thailand were analysed phylogenetically using three molecular markers (ITS, D1/D2 of 28S rDNA and ACT) and physiological races were determined on 10 differential rice cultivars. We found that S. oryzae is not the only causal agent of sheath rot in Thailand, but S. attenuatum was also found. Despite having similar morphological features, the phylogenetic analysis recognised 11 of 32 isolates as S. attenuatum and the remaining isolates as S. oryzae. This is the first report of S. attenuatum causing sheath rot of rice in Thailand in addition to S. oryzae. Evaluation of physiological races revealed high pathogenic diversity of the two species. Thus, 16 and 11 physiological races were recorded from 21 isolates of S. oryzae and 11 isolates of S. attenuatum, respectively. These results indicate that both S. oryzae and S. attenuatum are the causal agents of rice sheath rot and dirty panicle in Thailand and that they are pathologically diverse.

Gas-Phase Characterization of Redox-Active Manganese Complexes.

Manganese complexes exhibit a rich redox chemistry, usually accompanied by structural reorganization during the redox processes often followed by ligand dissociation or association. The push-pull ligand 2,6-diguanidylpyridine (dgpy) stabilizes manganese in the oxidation states +II, +III, and + IV in the complexes [Mn(dgpy)2]n+ (n = 2-4) without change in the coordination sphere in the condensed phase [Heinze et al., Inorganic Chemistry, 2022, 61, 14616]. In the condensed phase, the manganese(IV) complex is a very strong oxidant. In the present work, we investigate the stability and redox activity of the MnIV complex and its counterion (PF6-) adducts in the gas phase, using two modified 3D Paul ion trap mass spectrometers. Six different cationic species of the type [Mnx(dgpy)2(PF6)y]n+ (x = II, III, IV, y = 0-3, n = 1-3) could be observed for the three oxidation states MnIV, MnIII, and MnII, of which one observed complex also contains a reduced dgpy ligand. MnII species showed the highest relative stability in collision induced dissociation and UV/vis photo dissociation experiments. The lowest stability is observed in the presence of one or more counterions, which correlates to a lower total charge n+. Gas phase UV/vis spectra show similar features as the condensed phase spectra only differing in relative band intensities. The strongly oxidizing MnIV complex reacts with triethylamine (NEt3) in the gas phase to give MnIII, while MnIII species show little reactivity toward NEt3.

Fast Robust Point Cloud Registration Based on Compatibility Graph and Accelerated Guided Sampling

Point cloud registration is a crucial technique in photogrammetry, remote sensing, etc. A generalized 3D point cloud registration framework has been developed to estimate the optimal rigid transformation between two point clouds using 3D key point correspondences. However, challenges arise due to the uncertainty in 3D key point detection techniques and the similarity of local surface features. These factors often lead to feature descriptors establishing correspondences containing significant outliers. Current point cloud registration algorithms are typically hindered by these outliers, affecting both their efficiency and accuracy. In this paper, we propose a fast and robust point cloud registration method based on a compatibility graph and accelerated guided sampling. By constructing a compatible graph with correspondences, a minimum subset sampling method combining compatible edge sampling and compatible vertex sampling is proposed to reduce the influence of outliers on the estimation of the registration parameters. Additionally, an accelerated guided sampling strategy based on preference scores is presented, which effectively utilizes model parameters generated during the iterative process to guide the sampling toward inliers, thereby enhancing computational efficiency and the probability of estimating optimal parameters. Experiments are carried out on both synthetic and real-world data. The experimental results demonstrate that our proposed algorithm achieves a significant balance between registration accuracy and efficiency compared to state-of-the-art registration algorithms such as RANSIC and GROR. Even with up to 2000 initial correspondences and an outlier ratio of 99%, our algorithm achieves a minimum rotation error of 0.737° and a minimum translation error of 0.0201 m, completing the registration process within 1 s.

A model reduction framework of flow maldistribution and mitigation strategy for parallel flow field based polymer exchange membrane fuel cell

Mitigating flow maldistribution plays a crucial role in next-generation industry-used parallel-based flow field design for polymer exchange membrane fuel cells (PEMFCs), especially for larger-size cells and higher current density. In this study, a model reduction framework is proposed to alleviate flow maldistribution among parallel gas flow channels and associated computational burden. The gas flow features and electrochemical kinetics of four mitigation strategies including expanded manifold, partially porous media, dot matrix, and partially narrow structures are investigated in detail. The results indicate that the fuel cell model and a fluid flow model without the reactions show similar maldistribution features among parallel channels. The observed flow maldistribution can be effectively reduced by mitigating designs. Expanding the manifold space significantly alleviates the vortex impeding the gas entry to individual channel inlet, and the inlet exhibits better flow uniformity and pressure drop in the direction perpendicular to the channel than in the parallel to. The fully-inserted and manifold-inserted porous media methods present symmetrical quadratic-shaped flow patterns among channels, while that of the partially-inserted porous media inside channels is unsymmetrical. Additionally, the concentration loss with a partially narrowed design decreases by 44.7%, 12.9%, and 24.2% compared with the expanded manifold, partial porous media, and dot matrix designs at 1.8 A cm−2 owing to better flow uniformity and mass transport. These model framework and mitigation strategies are valuable for designing future high-performance industrial PEMFCs.

Unsupervised domain adaptive building semantic segmentation network by edge-enhanced contrastive learning

Unsupervised domain adaptation (UDA) is a weakly supervised learning technique that classifies images in the target domain when the source domain has labeled samples, and the target domain has unlabeled samples. Due to the complexity of imaging conditions and the content of remote sensing images, the use of UDA to accurately extract artificial features such as buildings from high-spatial-resolution (HSR) imagery is still challenging. In this study, we propose a new UDA method for building extraction, the contrastive domain adaptation network (CDANet), by utilizing adversarial learning and contrastive learning techniques. CDANet consists of a single multitask generator and dual discriminators. The generator employs a region and edge dual-branch structure that strengthens its edge extraction ability and is beneficial for the extraction of small and densely distributed buildings. The dual discriminators receive the region and edge prediction outputs and achieve multilevel adversarial learning. During adversarial training processing, CDANet aligns the cross-domain of similar pixel features in the embedding space by constructing the regional pixelwise contrastive loss. A self-training (ST) strategy based on pseudolabel generation is further utilized to address the target intradomain discrepancy. Comprehensive experiments are conducted to validate CDANet on three publicly accessible datasets, namely the WHU, Austin, and Massachusetts. Ablation experiments show that the generator network structure, contrastive loss and ST strategy all improve the building extraction accuracy. Method comparisons validate that CDANet achieves superior performance to several state-of-the-art methods, including AdaptSegNet, AdvEnt, IntraDA, FDANet and ADRS, in terms of F1 score and mIoU.

On the Response of the Transition Region and the Corona to Rapid Excursions in the Chromosphere

Spicules are the thin hair/grass-like structures that are prominently observed at the chromospheric solar limb. It is believed that fibrils and rapid blueshifted and redshifted excursions (RBEs and RREs; collectively referred to as REs) correspond to on-disk counterparts of type I spicules and type II spicules, respectively. Our investigation focuses on observing the response of these REs alongside similar spectral features in the chromosphere, transition region (TR), and corona, utilizing space–time plots derived from coordinated observations from the Swedish 1 m Solar Telescope/Hα, Interface Region Imaging Spectrograph (IRIS), and Solar Dynamics Observatory. Our analysis reveals upflowing REs, promptly reaching temperatures characteristic of the TR and corona, indicating a multithermal nature. Similarly, downflowing features exhibiting similar spectral signatures over the disk display plasma motion from the corona to chromospheric temperatures, demonstrating a multithermal nature. In addition to distinct upflows and downflows, we observe sequential upflow and downflow along the same path, depicting a distinctive parabolic trajectory in space–time plots of observations sampling TR and various coronal passbands. Similar to isolated upflows and downflows, these REs also exhibit a multithermal nature throughout their trajectory. Furthermore, our results reveal a more intricate motion of the REs in which both upflow and downflow coexist at the same spatial location. On a different note, our analysis, utilizing coordinated IRIS spectral observations, shows spatiotemporal redshifts/downflows in both the TR and chromosphere, suggesting that at least subsets of the strong redshifts/downflows observed in TR temperature spectra result from the return from the upper atmosphere flow of plasma in the form of bundles of spicules or features exhibiting similar spectra.

An efficient multi-level thresholding method for breast thermograms analysis based on an improved BWO algorithm

Breast cancer is a prevalent disease and the second leading cause of death in women globally. Various imaging techniques, including mammography, ultrasonography, X-ray, and magnetic resonance, are employed for detection. Thermography shows significant promise for early breast disease detection, offering advantages such as being non-ionizing, non-invasive, cost-effective, and providing real-time results. Medical image segmentation is crucial in image analysis, and this study introduces a thermographic image segmentation algorithm using the improved Black Widow Optimization Algorithm (IBWOA). While the standard BWOA is effective for complex optimization problems, it has issues with stagnation and balancing exploration and exploitation. The proposed method enhances exploration with Levy flights and improves exploitation with quasi-opposition-based learning. Comparing IBWOA with other algorithms like Harris Hawks Optimization (HHO), Linear Success-History based Adaptive Differential Evolution (LSHADE), and the whale optimization algorithm (WOA), sine cosine algorithm (SCA), and black widow optimization (BWO) using otsu and Kapur's entropy method. Results show IBWOA delivers superior performance in both qualitative and quantitative analyses including visual inspection and metrics such as fitness value, threshold values, peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and feature similarity index (FSIM). Experimental results demonstrate the outperformance of the proposed IBWOA, validating its effectiveness and superiority.

Predictors of ejection fraction recovery and baseline differences in patients with peripartum cardiomyopathy and dilated cardiomyopathy

Aims: Dilated cardiomyopathy (DCM) and peripartum cardiomyopathy (PPCM) are heart failure conditions with similar clinical, morphological and pathophysiological features but different underlying pathways. In PPCM and DCM patients, improvement in left ventricular ejection fraction (LVEF) varies depending on a number of factors. In this study, we aimed to determine the main differences between DCM and PPCM patients and the predictors of LVEF recovery in these patients. Methods: This cross-sectional, observational study included 33 consecutive female patients, 10 with PPCM and 23 with DCM, attending a tertiary cardiac center between March 2020 and April 2023. We performed a retrospective analysis of some clinical data and LVEF measurements. The main outcome was accepted as EF improvement at a follow-up of at least 12 months. Binary logistic analysis was conducted to assess predictive factors linked to LVEF recovery. This involved using binary logistic regression analysis to figure out odds ratio (OR) and 95% confidence interval (CI). Results: The PPCM group had a higher mean follow-up LVEF and LVEF value increase (p

Attention mechanism-enhanced graph convolutional neural network for unbalanced lithology identification

In this study, we propose a novel method for identifying lithology using an attention mechanism-enhanced graph convolutional neural network (AGCN). The aim of this method is to address the limitations of traditional approaches that evaluate unbalanced lithology by improving the identification of thin layers and small samples, while providing reliable data support for reservoir evaluation. To achieve this goal, we begin by using Principal Component Analysis (PCA) with maximum and minimum distance clustering (Max-min-distance) to correct the logging curves, which compensates for the low resolution of thin layers and enhances the accuracy of stratigraphic representation. Subsequently, we transform the logging data into graph-structured data by connecting distance similarity points and feature similarity points of the logging samples. We then use the graph convolutional network (GCN) to identify lithology, leveraging both labeled and unlabeled data to enhance the ability to identify lithology in small sample datasets. Additionally, our model incorporates a channel and spatial attention mechanism that assigns weights to the graph structure during lithology identification, improving the model’s capability to discern differences across samples. To evaluate the performance of our model, we constructed a lithology dataset comprising five wells and conducted experiments. The results indicate that our approach achieves a maximum accuracy of 97.67%, surpassing the performance of a singlestructure model in lithology identification. In conclusion, our proposed method provides a promising and effective approach for unbalanced lithology identification, significantly improving accuracy levels.

Exome sequencing revealed variants in SGCA and SIL1 genes underlying limb girdle muscular dystrophy and Marinesco-Sjögren syndrome patients.

Inherited neuromuscular (NMD) and neurodegenerative diseases (NDD) belong to two distinct categories that disturb different components of the nervous system, leading to a variety of different symptoms and clinical manifestations. Both NMD and NDD are a heterogeneous group of genetic conditions. Genetic variations in the SGCA and SIL1 genes have been implicated in causing Limb Girdle Muscular Dystrophy (LGMD), a type of neuromuscular disorder, and Marinesco-Sjögren Syndrome (MSS) which is a neurodegenerative disorder. In the present study, we have investigated four patients presenting LGMD and five patients with MSS features. After collecting detailed clinical and family history, necessary laboratory investigations, including estimation of a skeletal muscle marker enzyme serum creatine kinase (CK), nerve conduction study (NCS), electromyography (EMG), echocardiography (Echo), Magnetic resonance imaging (MRI -brain), CT-brain and X-rays were performed. Whole exome followed by Sanger sequencing was employed to search for the disease-causing variants. Physical examination in LGMD patients revealed poor muscle tone and facing difficulty in straightening up from the floor. Clinical history revealed frequent falls and strenuousness in climbing stairs. They started toe-walking in early childhood. Laboratory investigations confirmed elevated CK levels and abnormal NCS and EMG. The MSS patients showed abnormalities in gate and jerking movement, abnormal speech, and strabismus with cataract. MRI-brain showed cerebral atrophy in some MSS patients with elevated CK levels. Whole exome sequencing revealed a nonsense variant [c.C574T, p.(Arg192*)] in the SGCA gene and a frameshift [c.936dupG, p.(Leu313AlaFs*39)] in the SIL1 gene in LGMD and MSS patients, respectively. Our study emphasizes the significance of integrating clinical and genetic analyses for precise diagnosis and tailored management strategies in inherited NMD and NDD disorders. To the best of our knowledge, this is the first study documenting SGCA and SIL1 recurrent variants in subcontinent populations with few rare clinical features. The recurrent mutations expanding the global understanding of the mutation's geographic and ethnic distribution and contributing valuable epidemiological data. The study will facilitate genetic counseling for families experiencing similar clinical features, both within Pakistani populations and in other regions.

Segmental mandibulectomy and microvascular reconstruction with fibula free flap: Comparison between intraoral and extraoral approaches

BackgroundSegmental mandibulectomy and reconstruction of resulting defect can be performed via intraoral approach (IOA) or extraoral approach (EOA). Both approaches have advantages, disadvantages, indications, contra-indications to consider during their selection. ObjectiveTo compare IOA vs EOA of segmental mandibulectomy and microvascular reconstruction with fibula free flap (FFF). MethodsWe conducted a retrospective cohort study in which 51 patients who underwent segmental mandibulectomy and microvascular reconstruction with FFF from 2020 to 2024 were included, especially 17 patients by IOA and 34 patients by EOA, representing both groups of this study. Clinical characteristics, surgery parameters, and patients’ prognosis were evaluated. Patients’ satisfaction and DAS59 were assessed during follow-up. ResultsAmeloblastoma was the most frequent diagnosis (52.9% by IOA vs 70.6% by EOA); FFF was frequently positioned as double barrel (94.1% by IOA vs 88.2% by EOA). Compared to EOA group, IOA group had fewer intraoperative blood loss (MD = -112.2, 95% CI (-178.9 to -45.5), p = 0.001), higher satisfaction score (MD =1.3, 95% CI (0.9 to 1.7), p ˂ 0.001), and lower DAS59 score (MD = -0.5, 95% CI (-0.7 to -0.2), p ˂ 0.001). ConclusionBoth IOA and EOA were found safe and feasible, presenting similar perioperative features and postoperative outcomes. IOA patients were very satisfied with esthetic outcome than EOA patients. In absence of simultaneous immediate implant during mandibular FFF reconstruction, after stability of FFF on the defect site patients should always be referred to implantologist and/or prosthodontist for teeth restoration to improve functional and esthetic outcomes.

DMWMNet: A novel dual-branch multi-level convolutional network for high-performance mixed-type wafer map defect detection in semiconductor manufacturing

Wafer map defect detection plays an important role in semiconductor manufacturing by identifying root causes and accelerating process adjustments to ensure product quality and reduce unnecessary expenditures. However, existing methods have some limitations, such as low accuracy in mixed-type defect detection and poor recognition of similar defects and weak features. In this article, a novel dual-branch multi-level convolutional network (DMWMNet) is proposed for high-performance mixed-type wafer map defect detection. By fully considering the interrelationships between basic defects, defect number, and defect type, the network is designed to include two efficient parallel Branches and a Fusion classifier. Detecting defect types using basic defect discrimination and defect number detection is helpful for ameliorating problems with high complexity and low accuracy caused by multiple defect categories and feature overlaps. Furthermore, a composite loss function based on focal loss is employed to improve the network’s capacity to recognize weak features and similar defects. Experimental results on the MixedWM38 dataset show that DMWMNet has favorable mixed-type defect detection performance compared to other methods, with accuracy, precision, recall, F1 score, and MCC of 98.99%, 98.94%, 99.03%, 98.98%, and 98.97%, respectively.

Seed yield limitations of coriander (Coriandrum sativum L.) based on plant structure analysis

The present study describes the dependencies of key productivity and growth indicators of coriander plants on sowing rate and fertilization parameters. The aim of the research was to establish the patterns of coriander yield formation through the prism of plant structure considering modifications in nutrient conditions and sowing rates. Graphical models and correlation matrices were developed for three coriander varieties with similar morphological features (Oksanit, Nectar, Caribe). Data from field studies conducted in 2013–2015 in the Forest-Steppe zone of Ukraine serve as the basis for model construction. Two variable technological elements were studied (sowing rate of seeds; fertilization system) to determine the dependencies of biometric indicators for each variety. 3D graphical models of dependency for structural elements showed advantages over linear models. Seed yield of coriander in the study ranged from 1.24 t/ha with the lowest sowing rate to 1.91 t/ha with a sowing rate of 2.5 million seeds/ha. Maximum essential oil yield for cv. Oksanit was 46.4 kg/ha with a seed yield of 2.16 tons/ha. Cv. Nectar forms 45.6 kg/ha of essential oil with a seed yield of 2.01 tons/ha, and cv. Caribe forms 40.6 kg/ha with a seed yield of 2.06 tons/ha. Absence of statistically significant correlation between seed yield and essential oil output points to the peculiarities of the experiment. Research factors have opposing effects on the formation of essential oil, and consequently, its yield from the seed yield. The value of the results lies in the potential application of this approach in practice and the implementation of findings in the humid zone. The development of productivity models (like 3D plots) across a range of multiple factors simultaneously, based on crop structure indicators, allows for the prediction of target crop parameters, enabling the efficient utilization of available resources.

Centralised visual processing center for remote sensing target detection

Target detection in satellite images is an essential topic in the field of remote sensing and computer vision. Despite extensive research efforts, accurate and efficient target detection in remote sensing images remains unsolved due to the large target scale span, dense distribution, and overhead imaging and complex backgrounds, which result in high target feature similarity and serious occlusion. In order to address the above issues in a comprehensive manner, within this paper, we first propose a Centralised Visual Processing Center (CVPC), this structure is a parallel visual processing center for Transformer encoder and CNN, employing a lightweight encoder to capture broad, long-range interdependencies. Pixel-level Learning Center (PLC) module is used to establish pixel-level correlations and improve the depiction of detailed features. CVPC effectively improves the detection efficiency of remote sensing targets with high feature similarity and severe occlusion. Secondly, we propose a centralised feature cross-layer fusion pyramid structure to fuse the results with the CVPC in a top-down manner to enhance the detailed feature representation capability at each layer. Ultimately, we present a Context Enhanced Adaptive Sparse Convolutional Network (CEASC), which improves the accuracy while ensuring the detection efficiency. Based on the above modules, we designed and conducted a series of experiments. These experiments are conducted on three challenging public datasets, DOTA-v1.0, DIOR, and RSDO, showing that our proposed 3CNet achieves a more advanced detection accuracy while balancing the detection speed (78.62% mAP for DOTA-v1.0, 79.12% mAP for DIOR, and 95.50% mAP for RSOD).