Diverse Scales Research Articles

Application of correlation-based recurrent neural network in porosity prediction for petroleum exploration

Abstract Porosity prediction serves as a key metric to determine the characteristics of reservoirs and to reduce the exploration efforts, which in turn lead to enhancing the production planning and making economic vital decisions. In this sense, one can use low costly and low time-consumptive method by conducting experimental tests on various soil samples. Predicting porosity can be effectively utilized for identifying suitable drilling locations for petroleum exploration, thus reducing the need for extensive laboratory experiments beforehand. The seismic behavior has a direct impact on porosity features. This study works to find a mapping relationship between well-log porosity and seismic attributes. A Pearson correlation coefficient method has been proposed based on diverse scales to determine the most significant seismic attributes and to reduce data dimensionality. The recurrent prediction neural network models have been proposed for learning and prediction. Three neural network models are introduced, represented by Nonlinear Autoregressive (NARX) network, the Long Short-Term Memory (LSTM) network, and the Gated Recurrent Unit (GRU) network. All these proposed models are trained to associate the nonlinear relationship between the selected attributes and porosity. The seismic attributes are assigned as the inputs to the neural models, while the porosity represents the target of neural structure. Moreover, the correlated and uncorrelated strategies are involved to test the effective attributes based on seven scenarios with the proposed neural structures. The results showed that the correlated NARX network consistently delivers better prediction accuracy and minimum errors as compared to other models for different correlation scales. Consequently, correlated NARX network is the best choice for porosity prediction.

Multivariate Controls of Water–Carbon Coupling Relationship Under Various Land Use Types in the Thick Loess Deposits

ABSTRACTPrior studies have conducted extensive investigations of the water–carbon coupling relationship for aboveground vegetation and shallow soils, but the characteristics and multivariate controls in deep soils have not been fully explored. This has important implications for better understanding the water and carbon cycles of ecosystems. In this study, we attempted to examine the water–carbon coupling relationship in 18 m profiles under farmland (F), grassland (G), willow (SP), and poplar (P) in China's Loess Plateau. Specially, the multivariate controls of the water–carbon coupling relationship at diverse depths and scales were explored via the wavelet analysis. Individually, the conversion from F to G, SP, and P decreased the soil water storage within the 0–15 m by 553 (22% of F), 557 (22%), and 943 mm (38%). Land use change had little impact on soil organic or inorganic carbon. Even so, the poplar, compared with the willow, resulted in higher vertical variations in deep water and carbon. Furthermore, land use conversion increased the coherence between soil water and carbon in the whole layer. In comparison to F, which was dominated by water retention, G, SP, and P exhibited dominant carbon sequestration. The poplar reduced the root‐mean‐square deviation between soil water and organic carbon from 0.29 (F) to 0.13 in the 6‐ to 15‐m layer and between soil water and inorganic carbon from 0.32 (F) to 0.20 in the 0‐ to 2‐m layer. Sand content and electrical conductivity both played negative roles in maintaining water and carbon in the 6‐ to 15‐m and 0‐ to 15‐m layers at different scales, respectively. This study endeavors to present a novel viewpoint on land use management, with the potential to enhance ecosystem services in water‐limited and land‐degradation regions.

Enhancing preservation of ancient buildings through color imagery analysis and activation: take yuelu academy as an example

ABSTRACT The pivotal role of color imagery analysis and activation has been insufficiently explored in the realm of historic heritage conservation. This study addresses this issue by introducing a comprehensive methodological framework that encompasses acquisition, analysis, activation, and application of color information based on the NCD (Nippon Color & Design) color system, an innovative color semantic quantification theory. Utilizing the venerable Yuelu Academy (from A.D. 976) as a case study, color imagery analysis methods across diverse spatial scales have been summarized. Subsequently, these results are applied to activate the surrounding environment color, as well as its cultural souvenirs. The methodologies outlined herein hold profound significance in addressing challenges concerning visual color continuity in the preservation of historic architectural landscapes, offering innovative perspectives.

Exploring the factors shaping microscopic morphological traits: insights from the chrysophycean genus Synura (Stramenopiles)

ABSTRACT Synura represents a common freshwater protist genus known for producing morphologically diverse silica scales with species-specific secondary structures. In this study, we analysed 19 scale-bearing morphological traits across over 700 genetically verified records, spanning a broad environmental gradient in Europe. Focusing on Synura from section Petersenianae, we identified 29 species-level lineages, 13 of which had not been previously described. Interestingly, while species diversity increased northwards, morphological trait diversity exhibited the opposite trend, being higher in southern Europe. This suggests that abiotic factors play a significant role in shaping these traits, which are differently responsive to gradients than taxonomic diversity. We observed non-random distribution of scale morphological traits, influenced by abiotic factors such as temperature, precipitation and pH. Scale elongation appears to be an adaptation for living in oligotrophic waters with reduced sunlight availability, while the increasing distance between struts likely reflects an adaptation to low silica concentrations. Notably, specific morphological traits may provide more informative insights into environmental drivers than taxonomic units. The study sheds light on the intricate relationship between environmental conditions and the morphology of silica scales, emphasizing the importance of considering both taxonomic and morphological diversity in ecological research.

Geospatial Public Participation Tool Adoption in U.S. Natural Resource Agencies

Government agencies increasingly adopt geospatial online participation tools (geo-OPTs) for public engagement. Yet, little is known about the enablers and constraints surrounding their adoption. We draw from organizational innovation theory to address this knowledge gap related to the adoption of geo-OPTs by U.S. natural resource agencies. We conduct 35 interviews with natural resources agencies and triangulate our findings with policy documents. We find that enablers and constraints interact in distinctive ways that support and hinder innovation of geo-OPTs. Organizational champions navigate enablers and constraints within organizations and their broader environment to propel tool adoption and overcome structural barriers to public engagement in natural resources decision-making. By doing so, geo-OPTs become a pathway for agencies to overcome structural barriers and a tool for advancing human dimensions of natural resource management. These findings provide lessons applicable to public participation activities across diverse governance scales and issue areas.

An Improved iTransformer with RevIN and SSA for Greenhouse Soil Temperature Prediction

In contemporary agricultural practices, greenhouses serve as a critical component of infrastructure, where soil temperature plays a vital role in enhancing pest management and regulating crop growth. However, achieving precise greenhouse environmental control continues to pose a significant challenge. In this context, the present study proposes ReSSA-iTransformer, an advanced predictive model engineered to accurately forecast soil temperatures within greenhouses across diverse temporal scales, encompassing both long-term and short-term horizons. This model capitalizes on the iTransformer time-series forecasting framework and integrates Singular Spectrum Analysis (SSA) to decompose environmental variables, thereby augmenting the extraction of pivotal features, such as soil temperature. Furthermore, to mitigate the prevalent distribution shift issues inherent in time-series data, Reversible Instance Normalization (RevIN) is incorporated within the model architecture. ReSSA-iTransformer is adept at executing multi-step forecasts for both extended and immediate future intervals, thereby offering comprehensive predictive capabilities. Empirical evaluations substantiate that ReSSA-iTransformer surpasses conventional models, including LSTM, Informer, and Autoformer, across all assessed metrics. Specifically, it attained R2 coefficients of 98.51%, 97.03%, 97.26%, and 94.83%, alongside MAE values of 0.271, 0.501, 0.648, and 1.633 for predictions at 3 h, 6 h, 24 h, and 48 h intervals, respectively. These results highlight the model’s superior accuracy and robustness. Ultimately, ReSSA-iTransformer not only provides dependable soil temperature forecasts but also delivers actionable insights, thereby facilitating enhanced greenhouse management practices.

Instance Segmentation and 3D Pose Estimation of Tea Bud Leaves for Autonomous Harvesting Robots

In unstructured tea garden environments, accurate recognition and pose estimation of tea bud leaves are critical for autonomous harvesting robots. Due to variations in imaging distance, tea bud leaves exhibit diverse scale and pose characteristics in camera views, which significantly complicates the recognition and pose estimation process. This study proposes a method using an RGB-D camera for precise recognition and pose estimation of tea bud leaves. The approach first constructs an for tea bud leaves, followed by a dynamic weight estimation strategy to achieve adaptive pose estimation. Quantitative experiments demonstrate that the instance segmentation model achieves an mAP@50 of 92.0% for box detection and 91.9% for mask detection, improving by 3.2% and 3.4%, respectively, compared to the YOLOv8s-seg instance segmentation model. The pose estimation results indicate a maximum angular error of 7.76°, a mean angular error of 3.41°, a median angular error of 3.69°, and a median absolute deviation of 1.42°. The corresponding distance errors are 8.60 mm, 2.83 mm, 2.57 mm, and 0.81 mm, further confirming the accuracy and robustness of the proposed method. These results indicate that the proposed method can be applied in unstructured tea garden environments for non-destructive and precise harvesting with autonomous tea bud-leave harvesting robots.

A semi-supervised boundary segmentation network for remote sensing images

Accurately segmenting remote sensing images remains challenging due to the diverse target scales and ambiguous structural boundaries. In this work, we propose a semi-supervised boundary segmentation network (BS-GAN) to address these challenges. BS-GAN employs a semi-supervised learning approach to reduce dependency on labeled data while introducing a novel mixed attention (MA) mechanism to enhance segmentation accuracy by aggregating long-range contextual information. Additionally, we develop a Boundary Gating Module (BGM) to refine boundary segmentation through a multi-task learning strategy focused on boundary feature enhancement. Experimental results on three benchmark datasets demonstrate that BS-GAN achieves superior accuracy and generalization capabilities compared to existing segmentation networks.

LKAFFNet: A Novel Large-Kernel Attention Feature Fusion Network for Land Cover Segmentation

The accurate segmentation of land cover in high-resolution remote sensing imagery is crucial for applications such as urban planning, environmental monitoring, and disaster management. However, traditional convolutional neural networks (CNNs) struggle to balance fine-grained local detail with large-scale contextual information. To tackle these challenges, we combine large-kernel convolutions, attention mechanisms, and multi-scale feature fusion to form a novel LKAFFNet framework that introduces the following three key modules: LkResNet, which enhances feature extraction through parameterizable large-kernel convolutions; Large-Kernel Attention Aggregation (LKAA), integrating spatial and channel attention; and Channel Difference Features Shift Fusion (CDFSF), which enables efficient multi-scale feature fusion. Experimental comparisons demonstrate that LKAFFNet outperforms previous models on both the LandCover dataset and WHU Building dataset, particularly in cases with diverse scales. Specifically, it achieved a mIoU of 0.8155 on the LandCover dataset and 0.9326 on the WHU Building dataset. These findings suggest that LKAFFNet significantly improves land cover segmentation performance, offering a more effective tool for remote sensing applications.

An evaluation approach to PM2.5 policy effectiveness over South Korea based on a newly proposed scalable spatial decomposition method

Abstract Air quality management policies often exhibit spatial inconsistencies in effectiveness due to the diverse spatial scales of air pollution variability, which result from source characteristics as well as geographical and meteorological factors. To address this, the present study proposes a scalable spatial decomposition method to separate spatiotemporal air pollution data into background (nationwide), intercity-scale (tens of kilometers), and neighborhood-scale (several kilometers) components. This decomposition was achieved by introducing spatially varying effective ranges for intercity-scale variability at each station, based on the correlation coefficient distance of the background-removed component. Applying this approach to hourly fine particulate matter (PM2.5) concentrations from 535 monitoring stations across South Korea for 2021–2022, we evaluated the effectiveness of PM2.5 management policies. During the polluted cold season (December to March), the intercity-scale component contributed an average of approximately 18% of the total PM2.5 concentration in the Seoul Metropolitan Area (SMA) and Central Area (CA), which are densely populated and industrialized regions. In contrast, this component helped reduce PM2.5 levels in southeastern coastal areas, where high winds facilitate dispersion. The neighborhood-scale component contributed positively to PM2.5 levels near industrial complexes and ports but negatively in residential and commercial areas. The results demonstrate the effectiveness of central government-led intercity-scale regulations on total emissions allowances in the SMA and CA and highlight the need for additional local management targeting individual point sources near industrial complexes and ports. This study provides intuitive spatial decomposition tools for understanding PM2.5 pollution across spatial scales and offers policymakers a foundation for developing multi-scale mitigation strategies.

Systems Thinking as a Methodological Approach to Study Infrastructure Space in Architectural Design Education

In architectural education, urban-scale studies provide an opportunity for architectural students to study the challenges that cities confront and their physical and conceptual frameworks with a multidisciplinary approach. The design process necessitates the critical evaluation of the inputs that define, structure, and govern the cities and the acknowledgement of social, economic, ecological, geographical, and experiential conditions. The critical reading of the city also demands an understanding of its prevailing, speculative, and emergent conditions, which can be apprised through a cohesive structure of relations shaped by directives from various agents. Advocating for a novel methodological practice in architectural education, this approach fosters the engagement of architecture students with the networks, constellations, and associations of contemporary urban conditions. With this conceptual framework, the paper speculates on the potential of introducing systems thinking as a methodology for architectural education, which encourages the study of interrelations between different parties, in diverse scales, to design contemporary urban conditions. It subjects students’ works in the fourth-year architectural design studio, where systems thinking is acknowledged as a methodology to study the notion of infrastructure space. In these studies, infrastructure space is considered as the site of multiplicities, coexistences, and overlaps beyond its typical association with “physical networks for transportation, communication or utilities” (Easterling, 2014). Studying the infrastructure space through a systems thinking approach is believed to enable the integration of inchoate states and territories of local, trans-local, and global occurrences. To sum up, the paper will discuss the outputs of integrating systems thinking in architectural education, and the reconceptualization of ‘infrastructure space’ as an instrumental approach in dealing with the complex structure of cities.

Scale dependence of bird diversity in London

ContextUnderstanding drivers of biodiversity in cities can be mutually beneficial for ecosystems and people. Crowd-sourced bird observations provide an opportunity to assess how patterns of bird diversity change across observation scales and suggest driving processes.ObjectivesWe assessed the scale dependence of bird diversity within a 128 × 128 km extent over London’s urban–rural gradient to suggest scales at which key drivers may be operating.MethodsWe quantified scale variance of bird diversity across scales from 500 m to 64,000 m for three groups of species (All, Passeriformes, and Anseriformes and Charadriiformes combined). We estimated diversity by aggregating observations into a series of grids and computed comparable diversity estimates within each cell using interpolation and rarefaction. We calculated the variance explained by each scale for common diversity metrics.ResultsThe results show that bird diversity patterns around London vary by scale, and that the location of high variance shifts across the study area depending on both scale and species group. The variance of Passeriformes diversity gradually shifted from the urban core to the periphery, while variance of Anseriformes and Charadriiformes diversity occurred near water features.ConclusionsThe results suggest that the urban–rural gradient and location of water are two properties of the study extent around London influencing the scale dependance of bird diversity that could be used to ground scale considerations of further modeling efforts.

An Evaluation of the Performance of Remote Sensing Indices as an Indication of Spatial Variability and Vegetation Diversity in Alpine Grassland

Vegetation diversity is a crucial indicator for evaluating grassland ecosystems. Remote sensing technology has great potential in assessing grassland vegetation diversity. In this study, the relationship between remote sensing indices and species diversity was investigated at varying spatial and temporal scales in Bayanbulak Grassland National Nature Reserve, China. Spectral variation, defined as the coefficient of variation in vegetation indices, was used as a proxy for species diversity, which was quantified using species diversity indices. The “spectral diversity-species diversity” relationship was validated across diverse spatial scales and between different years using Sentinel-2 images and ground investigation data. This study found that Kendall’s τ coefficients showed the best performance in evaluating the relationship between the coefficient of variation in VIs (CVVIs) and species diversity index. The highest τ value was observed for CVNDVI in 2017 (τ = 0.660, p < 0.01), followed by the Shannon index in 2018 (τ = 0.451, p < 0.01). In addition, CVEVI demonstrated a significant positive correlation with the Shannon-Wiener Index at the 50 m scale (τ = 0.542), and the highest relationship τ between CVNDVI and the Shannon-Wiener Index was observed at the 100 m scale (τ = 0.660). The Shannon-Wiener Index in relation to CVVIs performs better in representing changes in grassland vegetation. Spatial scales and vegetation indices influence the assessment of grassland vegetation diversity. These findings underscore the critical role of remote sensing technology in assessing grassland vegetation diversity across various scales, offering valuable support tools for measuring regional grassland vegetation diversity.

The Spatial Distribution of Brain Metastasis Is Determined by the Heterogeneity of the Brain Microenvironment.

It is now understood that brain metastases do not occur randomly but have distinct spatial patterns depending on the origin of the cancer. According to the "seed and soil" hypothesis, the final colonization of metastatic cells is the result of their adaptation to the altered environment. To investigate the most favorable microenvironment for brain metastasis, we analyzed neuroimaging data from 177 patients with breast cancer brain metastasis and 548 patients with lung cancer brain metastasis to create a replicable probabilistic map of metastatic locations. Additionally, we used population-based data from open repositories to generate brain atlases of diverse microenvironment features, including gene expression, functional connectivity, glucose metabolism, and neurotransmitter transporters/receptors. We then compared the spatial correlation between brain metastasis frequency and these features, after which we constructed a general linear model to identify the most significant variables that contributed to tumor location predilection. Our findings revealed that brain metastases from breast cancer and lung cancer had distinct radiographic characteristics and distribution patterns. Breast cancer tended to metastasize in brain regions with decreased expression of genes associated with immunity and metabolism and reduced levels of connectomic hubness and glucose metabolism. In contrast, lung cancer had a higher probability of metastasizing in regions with active metabolism. Moreover, neurotransmitter systems play various roles in determining tumor location. These results provide new insights into the adaptation of metastatic cells to the brain microenvironment and illustrate how factors on diverse biological scales can affect the colonization of brain metastases.

Volumetric medical image segmentation via fully 3D adaptation of Segment Anything Model

The Segment Anything Model (SAM) exhibits exceptional generalization capabilities in diverse domains, owing to its interactive learning mechanism designed for precise image segmentation. However, applying SAM to out-of-distribution domains, especially in 3D medical image segmentation, poses challenges. Existing methods for adapting 2D segmentation models to 3D medical data treat 3D volumes as a mere stack of 2D slices. The essential inter-slice information, which is pivotal to faithful 3D medical image segmentation tasks, is unfortunately neglected. In this work, we present the 3D Medical SAM-Adapter (3DMedSAM), a pioneer cross-dimensional adaptation, leveraging SAM’s pre-trained knowledge while accommodating the unique characteristics of 3D medical data. Firstly, to bridge the dimensional gap from 2D to 3D, we design a novel module to replace SAM’s patch embedding, ensuring a seamless transition into 3D image processing and recognition. Besides, we incorporate a 3D Adapter while maintaining the majority of pre-training parameters frozen, enriching deep features with abundant 3D spatial information and achieving efficient fine-tuning. Given the diverse scales of anomalies present in medical images, we also devised a multi-scale 3D mask decoder to elevate the network’s proficiency in medical image segmentation. Through various experiments, we showcase the effectiveness of 3DMedSAM in achieving accurate and robust 3D segmentation on both single-target segmentation and multi-organ segmentation tasks, surpassing the limitations of current methods.

A Novel Framework for Heterogeneity Decomposition and Mechanism Inference in Spatiotemporal Evolution of Groundwater Storage: Case Study in the North China Plain

AbstractProperly understanding the evolution mechanisms of groundwater storage anomaly (GWSA) is the basis of making effective groundwater management strategies. However, current analysis methods cannot objectively capture the spatiotemporal evolution characteristics of GWSA, which might lead to erroneous inferences of the evolution mechanisms. Here, we developed a new framework to address the challenge of spatiotemporal heterogeneity in the GWSA evolution analysis. It is achieved by integrating the Bayesian Estimator of Abrupt change, Seasonal change, and Trend (BEAST), the Balanced Iterative Reducing and Clustering using Hierarchies (BIRCH), and the Optimal Parameters‐based Geographical Detector (OPGD). In the case study of the North China Plain (NCP), the GWSA time series is divided into four stages by three trend change points in BEAST. An increasing trend of GWSA is observed at Stage IV, and the third trend change point occurs before the third seasonal change point. This distinguishes the positive feedback of anthropogenic interventions and the effects of seasonal precipitations for the first time. Moreover, the spatial distribution of GWSA in the NCP is classified into two clusters by BIRCH in each stage. The differences in GWSA trends and responses to environmental changes between Cluster‐1 and Cluster‐2 are significant. Then the driving effects of 16 factors on the evolution of GWSA are identified using OPGD, in which the contributions of topographic and aquifer characteristics are highlighted by quantitative analysis. This framework provides a novel method for examining the spatiotemporal heterogeneity of GWSA, which can be extended to analyze spatiotemporal trends in GWSA at diverse scales.

AttentionEP: Predicting Essential Proteins via Fusion of Multiscale Features by Attention Mechanisms

Identifying essential proteins is of utmost importance in the field of biomedical research due to their essential functions in cellular activities and their involvement in mechanisms related to diseases. In this research, a novel approach called AttentionEP for predicting essential proteins (EP) is introduced by attention mechanisms. This method leverages both cross-attention and self-attention frameworks, focusing on enhancing prediction accuracy through the integration of features across diverse scales. Spatial characteristics of proteins are obtained from the protein-protein interaction (PPI) network by employing Graph Convolutional Networks (GCN) and Graph Attention Networks (GAT). Following this, Bidirectional Long Short-Term Memory networks (BiLSTM) are employed to derive temporal features from gene expression datasets. Furthermore, spatial characteristics are derived by integrating data on subcellular localization with the application of Deep Neural Networks (DNN). In order to effectively integrate features across multiple scales, initial steps involve the application of self-attention techniques to derive essential insights from each unique data set. Following this, mechanisms involving self-attention and cross-attention are employed to enhance the interaction between diverse information sources. To identify essential proteins, a classifier based on the ResNet architecture is developed. The findings from the experiments indicate that the method introduced here shows superior performance in identifying essential proteins, recording an Area Under the Curve (AUC) value of 0.9433. This approach shows a considerable advantage over established techniques. The findings of this study provide a significant advancement in the comprehension of critical proteins, revealing promising potential for applications in the development of therapeutics and addressing various diseases.

Atomic Force Microscopy: Mechanosensor and Mechanotransducer for Probing Biological System from Molecules to Tissues.

Atomic Force Microscopy (AFM) is a powerful technique with widespread applications in various scientific fields, including biology. It operates by precisely detecting the interaction between a sharp tip and a sample surface, providing high-resolution topographical information and mechanical properties at a nanoscale. Through the years, a deeper understanding of this tip-sample interaction and the mechanisms by which it can be more precisely regulated have invariably led to improvements in AFM imaging. Additionally, AFM can serve not only as a sensor but also as a tool for actively manipulating the mechanical properties of biological systems. By applying controlled forces to the sample surface, AFM allows for a deeper understanding of mechanotransduction pathways, the intricate signaling cascades that convert physical cues into biochemical responses. This review, is an extensive overview of the current status of AFM working either as a mechanosensor or a mechanotransducer to probe biological systems across diverse scales, from individual molecules to entire tissues is presented. Challenges are discussed and potential future research directions are elaborated.

Commoning practices: the political significance of facts

ABSTRACT Within this cluster, which beckons a collective exploration of facts, this contribution hinges upon two overarching discussions. Firstly, it underscores that facts are not self-evident entities independent of human experiences, challenging conventional notions of natural environmental causality. The second discussion posits that the social life and significance of facts in Latin America are determined by what facts do or can potentially do. I explore how socio-environmental entanglements transform into certainties and convictions – facts that either permit or obstruct political decisions regarding collective environmental actions. Inquiring into the potential actions of facts is an integral aspect of civic epistemology, as defined in science studies, which traces “the institutionalized practices by which members of a given society test and deploy knowledge claims used as a basis for making collective choices” [Jasanoff (2005). Designs on Nature: Science and Democracy in Europe and the United States. Princeton, NJ: Princeton University Press, 255]. My contribution to this debate lies in elucidating how these institutionalized practices are ingrained across diverse social scales, extending far beyond the realms of technoscientific scenarios and state institutions, within the fabric of everyday commoning practices.

Learning to zoom: Exploiting mixed-scale contextual information for object detection

With the development of deep neural networks, object detection has made a significant leap. However, apart from intrinsic intra and inter class differences, the objects in real-world scenarios may encounter diverse scales, blurred appearance or even severe occlusion. Inspired by the human behavior of looking at blurred images, i.e., zooming in/out, we propose a Mixed-Scale Network (dubbed for MSNet) to tackle these issues. Specifically, MSNet employs the zoom strategy to exploit mixed-scale contextual information, which fully unleashes the representation ability of deep neural networks. Firstly, the global feature aggregation module and global feature enhancement module aims at aggregating mixed-scale features from the global perspective, which learns the transform offset of pixels to align the higher-res features contextually. Moreover, the local feature aggregation module enriches the instance-level feature by adaptively aligning the features of different receptive fields. Extensive experiments on MS COCO demonstrated the effectiveness of MSNet, yielding significant improvements of 0.7–2.8 points in APbox over baselines when paired with different detectors, backbones, and schedules. In addition, for pixel-level prediction tasks including semantic segmentation and instance segmentation, the proposed method gains consistent improvements of 0.6–2.7 points in mIoU/APmask over baselines, which substantiates the effectiveness of MSNet further.