Scale Variations Research Articles

Underwater target detection algorithm based on feature enhancement and feature fusion

Abstract In the field of target detection, identifying underwater targets remains a formidable challenge due to the complexity of the underwater environment and the substantial variations in target scale. To tackle the difficulties of extracting complex features in such conditions, this study introduces a novel feature extraction module built upon the YOLOv7 model. The module incorporates a custom-designed information attention mechanism and deformable convolution, enhancing the network’s ability to capture and process intricate underwater features. Additionally, a feature enhancement and interaction module was developed further to strengthen the extraction and utilization of key features. An adaptive feature fusion module was also introduced, allowing for dynamic adjustment of fusion weights based on the characteristics of the input features, thereby enabling more efficient and accurate multi-scale feature fusion. Experimental results reveal that, compared to the original YOLOv7 model on the URPC2020 dataset, the proposed method achieves improvements of 3.3% and 2.7% in mAP50 and mAP50:95 metrics, respectively. These significant performance gains highlight the effectiveness of the proposed approach in underwater target detection and provide valuable new directions for research and applications in this domain.

Using ESPEN data for evidence-based control of neglected tropical diseases in sub-Saharan Africa: A comprehensive model-based geostatistical analysis of soil-transmitted helminths.

The Expanded Special Project for the Elimination of Neglected Tropical Diseases (ESPEN) was launched in 2019 by the World Health Organization and African nations to combat Neglected Tropical Diseases (NTDs), including Soil-transmitted helminths (STH), which still affect over 1.5 billion people globally. In this study, we present a comprehensive geostatistical analysis of publicly available STH survey data from ESPEN to delineate inter-country disparities in STH prevalence and its environmental drivers while highlighting the strengths and limitations that arise from the use of the ESPEN data. To achieve this, we also propose the use of calibration validation methods to assess the suitability of geostatistical models for disease mapping at the national scale. We analysed the most recent survey data with at least 50 geo-referenced observations, and modelled each STH species data (hookworm, roundworm, whipworm) separately. Binomial geostatistical models were developed for each country, exploring associations between STH and environmental covariates, and were validated using the non-randomized probability integral transform. We produced pixel-, subnational-, and country-level prevalence maps for successfully calibrated countries. All the results were made publicly available through an R Shiny application. Among 35 countries with STH data that met our inclusion criteria, the reported data years ranged from 2004 to 2018. Models from 25 countries were found to be well-calibrated. Spatial patterns exhibited significant variation in STH species distribution and heterogeneity in spatial correlation scale (1.14 km to 3,027.44 km) and residual spatial variation variance across countries. This study highlights the utility of ESPEN data in assessing spatial variations in STH prevalence across countries using model-based geostatistics. Despite the challenges posed by data sparsity which limit the application of geostatistical models, the insights gained remain crucial for directing focused interventions and shaping future STH assessment strategies within national control programs.

Effects of Topography on Vegetation Coverage Changes in Rural Areas Under Multi-scale Scenarios

Revealing the characteristics of vegetation coverage change and the effects of topography on vegetation coverage change in rural areas since the reform and opening up has direct implications for further understanding of human-land coupling processes and resource and environmental changes and provides a reference for ecological environment protection in rural revitalization. Using the Taihua Town in Yixing City, Jiangsu Province, as a rural case study, we revealed the basic features of rural vegetation coverage change from 1986 to 2020. Furthermore, using 0.25 km2（scenario 1）, 0.50 km2（scenario 2）, and 1.00 km2 （scenario 3） as scale scenarios in the study area, both the single and comprehensive topographic effects on vegetation coverage change were explored. The study produced several important results： ① Vegetation coverage in the study area generally maintained an upward trend from 1986 to 2020, during which the spatial pattern of vegetation coverage distribution was generally high in the southeast and southwest and low in the north （northwest）. During this period, the vegetation coverage in the study area increased from 74.15% to 83.44%, and the spatial pattern of vegetation coverage distribution generally remained similar. ② Under the multi-scale scenarios, the vegetation coverage distribution in the study area generally maintained a similar spatial pattern from 1986 to 2020, excepting a few local differences. Under the same-scale scenario, Moran's index of vegetation coverage in the study area was higher in 2020 than in 1986. Taking scenario 1 as an example, Moran's index of vegetation coverage in the study area increased from 0.603 to 0.652 between 1986 and 2020. ③ Under the multi-scale scenarios, from 1986 to 2020, vegetation coverage in the study area generally maintained an upward trend with increase of elevation, slope, and topographic relief. As an example of the vegetation coverage change with elevation, in scenario 1 in 2020, the vegetation coverage in the study area increased from 64.50% in the area with elevations of 0-50 m to 98.77% in the area with elevations above 350 m. ④ Under the multi-scale scenarios, the comprehensive topographic effects on mean vegetation coverage change in the study area from 1986 to 2020 showed that elevation, slope, and topographic relief all had positive effects, and their effects were greater in scenario 3 than in scenario 2 and then in scenario 1 for the same topographic factor. According to the comprehensive topographic effects on mean vegetation coverage change in the study area from 1986 to 2020, the equation simulation coefficients of elevation under scenarios 1, 2, and 3 were 0.271, 0.281, and 0.283, respectively. The respective slope values were 0.283, 0.289, and 0.294, and the topographic relief values were 0.283, 0.292, and 0.295. This suggests that the scale variation has impacts on the comprehensive topographic effects. The results of this study are valuable for understanding the changes in the rural ecological environment under multi-scale scenarios since the reform and opening up.

Combining Two Distinct Methods to Resolve Spatial Variation in Attenuation and Earthquake Source Parameters

ABSTRACT Stress drop is a fundamental parameter in ground-motion modeling and seismic hazard assessment, but spectral estimates are subject to considerable uncertainties. A variety of factors cause different methods to yield different results, including the complexity of the seismic source, the assumptions inherent in the models used, the limited range of frequencies available, and the inherent difficulty in removing the propagation effects along the wave path. A primary challenge is determining whether the observed variations in spectral stress-drop estimates represent characteristics of the seismic source or the propagation path. We compare the performance of two methods applied to the 2019 Ridgecrest, California, earthquake sequence, each of which addresses the trade-offs between propagation and source in different ways. The first method, referred to as the spectral-fitting approach, operates on the hypothesis that the path effects remain constant across the spatial and temporal range of the sources under investigation. This approach assumes a level of uniformity in the propagation effects that simplifies the analysis. The second method, referred to as the spectral ratio approach, is based on the hypothesis that a small, collocated event will experience identical propagation effects to the earthquake of interest, potentially accounting for smaller scale variation in propagation effects. Our comparison reveals that the choice of method is not only influenced by the specifics of the data and the seismic events but also significantly constrained by the geological heterogeneity and consequent spatial variability of site and propagation effects in the study area. If an approach involves assuming a homogeneous attenuation structure, any spatial variation in attenuation structure will lead to this variation being incorrectly mapped into apparent source stress-drop variations. Understanding the local geology and structural heterogeneity, combined with using methods with contrasting underlying assumptions are good approaches to improving the reliability of estimated spectral stress drops.

Variations in the radiation intensity of the pulsar B0950+08: 9 years of monitoring at a frequency of 110 MHz

The analysis of variations in the radiation intensity of the pulsar B0950+08 from 2014 to 2022 with scales from minutes to years was carried out. The observations were obtained in a round-the-clock daily survey conducted on the Large Phased Array (LPA) radio telescope. The high variability of radiation is shown not only from pulse to pulse, but also at scales greater than 3 min. The average value of the estimated amplitude of these variations in 3.2 minutes is 25 Jy, the modulation index is 1. The average relative amplitude of the interpulse (IP) is 2.00 ± 0.28 % of the main pulse. In individual pulses, the amplitude of the interpulse may exceed the amplitude of the main pulse (MP), but this is a rare event. Radiation is observed in almost the entire period of the pulsar. For the first time, the relative amplitude of radiation between the main pulse and the interpulse (radiation bridge) was measured. When averaging about 10 hours, it varies from 0.8% to 1.31% with an average value of 1.04 ± 0.28 %. A high correlation was found between MP and IP amplitude variations both when averaging profiles over 3.2 minutes and when averaging over years. This correlation is due to refractive interstellar scintillation. The frequency scale of IP diffraction interstellar scintillation was measured for the first time and it was shown that the spectral forms for IP and MP are well correlated and have the same frequency scale. There are strong variations in the frequency scale of scintillation f dif from session to session (time interval from one day) on scales of 200–800 kHz. The refractive scale of scintillation for 1–2 days has been determined. A modulation of radiation with a characteristic scale of about 130 days was detected, which, apparently, is also associated with refractive scintillation.

DeepMoIC: multi-omics data integration via deep graph convolutional networks for cancer subtype classification

BackgroundAchieving precise cancer subtype classification is imperative for effective prognosis and treatment. Multi-omics studies, encompassing diverse data modalities, have emerged as powerful tools for unraveling the complexities of cancer. However, owing to the intricacies of biological data, multi-omics datasets generally show variations in data types, scales, and distributions. These intractable problems lead to challenges in exploring intact representations from heterogeneous data, which often result in inaccuracies in multi-omics information analysis.ResultsTo address the challenges of multi-omics research, our approach DeepMoIC presents a novel framework derived from deep Graph Convolutional Network (GCN). Leveraging autoencoder modules, DeepMoIC extracts compact representations from omics data and incorporates a patient similarity network through the similarity network fusion algorithm. To handle non-Euclidean data and explore high-order omics information effectively, we design a Deep GCN module with two strategies: residual connection and identity mapping. With extracted higher-order representations, our approach consistently outperforms state-of-the-art models on a pan-cancer dataset and 3 cancer subtype datasets.ConclusionThe introduction of Deep GCN shows encouraging performance in terms of supervised multi-omics feature learning, offering promising insights for precision medicine in cancer research. DeepMoIC can potentially be an important tool in the field of cancer subtype classification because of its capacity to handle complex multi-omics data and produce reliable classification findings.

Computer-Vision-Based Product Quality Inspection and Novel Counting System

In this study, we aimed to enhance the accuracy of product quality inspection and counting in the manufacturing process by integrating image processing and human body detection algorithms. We employed the SIFT algorithm combined with traditional image comparison metrics such as SSIM, PSNR, and MSE to develop a defect detection system that is robust against variations in rotation and scale. Additionally, the YOLOv8 Pose algorithm was used to detect and correct errors in product counting caused by human interference on the load cell in real time. By applying the image differencing technique, we accurately calculated the unit weight of products and determined their total count. In our experiments conducted on products weighing over 1 kg, we achieved a high accuracy of 99.268%. The integration of our algorithms with the load-cell-based counting system demonstrates reliable real-time quality inspection and automated counting in manufacturing environments.

Mapping typical LULC classes using spatiotemporal analysis and the thresholds of spectral optical satellite imagery indices: a case study in Algiers city.

Land use and land cover (LULC) dynamics have a substantial impact on human-environment interactions. Nowadays, remote sensing imagery has emerged as a useful tool for mapping and tracking LULC changes. Spectral optical indices derived from remote sensing data can provide insightful information about vegetation health, urban expansion, water bodies, deforestation patterns, and many other applications. The present study examines the use of popular optical spectral indices: vegetation index (NDVI), water indices (NDWI and MNDWI), urban indices (UI and NDBI), and bare land index (MNDBI) in threshold-based classification for LULC mapping using Algiers (Algeria) as a case study, and assesses the potential impacts of their spatiotemporal (at a seasonal and annual temporal scales) variations associated with natural seasonal changes and/or the evolution of the city's fabric. Here, a geo-statistical analysis of the values of the selected spectral indices at the level of each LU-class is conducted, threshold values (that account for seasonal variations) are proposed, and a classification approach (making use of best performing indices) is proposed and tested. Although fast and easy to implement, the proposed threshold-based LULC classification approach was successfully used for mapping LULC for the study zone with a high accuracy (an overall accuracy of 90.20 and a kappa of 0.84 for the demonstration year of 2017). The outcomes of the study heighten the potential and the limitations of the use of spectral indices for LULC mapping practices and consequent applications in environmental and urban studies.

A coarse-to-fine image registration method based on autocorrelation structural difference information

ABSTRACT The automatic registration of synthetic aperture radar (SAR) and optical images is still a challenging problem due to different imaging mechanisms. This letter proposes a coarse-to-fine image registration method that leverages self-similarity structural difference information. In the coarse registration stage, a Scale-Invariant Feature Transformation-based (SIFT-like) method is employed, complemented by an improved Fast Sample Consensus (IFSC) method to eliminate mismatched point pairs by probabilistic and geometric information. This stage ensures robustness against scale and rotational variations. In the fine registration stage, robust feature points are selected by utilizing phase and edge structural information. A descriptor which based on phase consistency and autocorrelation structural difference (ASDPC) is constructed to capture the structural variations between region blocks, and a fine search is carried out within the neighbourhood of the already matched feature points, so as to find more accurate matched feature points and obtain fine registration. The experimental results demonstrate that the proposed method provides robust and accurate registration for optical-to-SAR images.

A hybrid transformer masked time-domain denoising network for vibration signals

Abstract Health-condition-sensitive vibration information is prone to be swamped by widespread noise. Denoising is always indispensable, but existing methods still lack adaptability. Therefore, a novel intelligent denoising framework called a hybrid transformer masked time-domain denoising network (HTMTDN) is proposed. First, a dense dilation convolution block and a hybrid transformer are constructed to deal with fault impulse scale variations and unexpected noise frequency bands respectively, which greatly improves the adaptive denoising capability and relieves tough denoising parameter tuning. Further, interpretable time- and frequency-domain joint constraints are constructed to enhance the network’s optimization ability under strong noise. Finally, a novel strategy called overlapping reconstruction is introduced to recover 1D signals from 2D signal segments. Extensive experiments based on two bearing fault datasets with variable loads and rotation speeds confirm the remarkable performance of the HTMTDN under low signalnoise ratios, and present good adaptability in 15 health conditions without separate hyperparameter tuning, which shows promise for real-world applications.

Appendage Sizes in Three Rodent Species are Affected Indirectly by Climate Factors via Their Effects on Body Size.

In mammals, temporal and spatial variation in appendage sizes within and among species may be driven by variations in ambient temperature and allometric scaling. Here, we use two decades of morphological data on three rodent species distributed across vast latitudinal gradients in China to estimate temporal and spatial trends of tail, hind-foot, and ear lengths. Further, we test 14 climate variables to identify the critical drivers of these trends and use structural equation modeling (SEM) to analyze whether the effects of climate variables on the appendage lengths are direct or indirect, via effects on body length. Relative to body length, and in contradiction to Allen's rule, all appendage lengths remained unchanged over time and across space. By contrast, absolute appendage lengths increased in one species (Apodemus agrarius) over time and in two species (A. agrarius and Rattus norvegicus) across space; and most of the appendage lengths in the two species were associated with annual mean minimum temperature in the year preceding capture (PreAnnMinTemp). The SEM results suggest that PreAnnMinTemp affected absolute appendage lengths indirectly through body length. In addition, except for tail length in two species and both hind-foot and ear length in one species, absolute appendage lengths scaled allometrically with body length. These results suggest that the distinct temperature-appendage-length patterns among and within species arise from species-specific temperature sensitivities and appendage-specific ontogenetic rates and functions.

Sub‐centennial upper water column structure variability of the tropical Indian Ocean since the Last Glacial Maximum

ABSTRACTThe upper water column (UWC) is involved in the worldwide distribution of heat and nutrients. However, global warming is expected to alter the UWC structure due to temperature‐ and precipitation‐induced stratification. This impact of global warming is comparatively more pronounced in the Indian Ocean. Therefore, to understand the fate of the UWC structure in the future, it is important to comprehend its past variability in the Indian Ocean. To achieve this, we have reconstructed sub‐centennial scale variations in the UWC structure of the tropical central Indian Ocean by using the isotopic composition (δ18O and δ13C) of surface‐dwelling planktic foraminifera Globigerinoides ruber and subsurface‐dwelling Neogloboquadrina dutertrei. The difference in the isotopic composition of the surface‐ and subsurface‐dwelling species (Δδ18Or‐d and Δδ13Cr‐d) was used to understand the UWC structure variability in the tropical central Indian Ocean. We report a deeper mixed layer and thermocline during the last glacial period, followed by a shoaling of the thermocline during the glacial–interglacial transition. The thermocline also deepened during intense boreal summer monsoon events such as in the Bølling–Allerød and early Holocene. Our findings suggest that the tropical Indian Ocean UWC is influenced primarily by the intensity of the monsoon winds. From the comparison of our record with those of the eastern and western Indian Ocean, it is evident that a similar climatic forcing governed the central and eastern regions during the last glacial–interglacial period.

Crowd Counting Based on Multiscale Spatial Guided Perception Aggregation Network.

Crowd counting has received extensive attention in the field of computer vision, and methods based on deep convolutional neural networks (CNNs) have made great progress in this task. However, challenges such as scale variation, nonuniform distribution, complex background, and occlusion in crowded scenes hinder the performance of these networks in crowd counting. In order to overcome these challenges, this article proposes a multiscale spatial guidance perception aggregation network (MGANet) to achieve efficient and accurate crowd counting. MGANet consists of three parts: multiscale feature extraction network (MFEN), spatial guidance network (SGN), and attention fusion network (AFN). Specifically, to alleviate the scale variation problem in crowded scenes, MFEN is introduced to enhance the scale adaptability and effectively capture multiscale features in scenes with drastic scale variation. To address the challenges of nonuniform distribution and complex background in population, an SGN is proposed. The SGN includes two parts: the spatial context network (SCN) and the guidance perception network (GPN). SCN is used to capture the detailed semantic information between the multiscale feature positions extracted by MFEN, and improve the ability of deep structured information exploration. At the same time, the dependence relationship between the spatial remote context is established to enhance the receptive field. GPN is used to enhance the information exchange between channels and guide the network to select appropriate multiscale features and spatial context semantic features. AFN is used to adaptively measure the importance of the above different features, and obtain accurate and effective feature representations from them. In addition, this article proposes a novel region-adaptive loss function, which optimizes the regions with large recognition errors in the image, and alleviates the inconsistency between the training target and the evaluation metric. In order to evaluate the performance of the proposed method, extensive experiments were carried out on challenging benchmarks including ShanghaiTech Part A and Part B, UCF-CC-50, UCF-QNRF, and JHU-CROWD ++ . Experimental results show that the proposed method has good performance on all four datasets. Especially on ShanghaiTech Part A and Part B, CUCF-QNRF, and JHU-CROWD ++ datasets, compared with the state-of-the-art methods, our proposed method achieves superior recognition performance and better robustness.

Bycatch of northern fulmar (Fulmarus glacialis) in Norwegian longline fisheries: Assessing spatiotemporal variations in scale and risk to improve management

Bycatch of northern fulmar (Fulmarus glacialis) in Norwegian longline fisheries: Assessing spatiotemporal variations in scale and risk to improve management

AMS-Net: Modeling Adaptive Multi-Granularity Spatio-Temporal Cues for Video Action Recognition.

Effective spatio-temporal modeling as a core of video representation learning is challenged by complex scale variations in spatio-temporal cues in videos, especially different visual tempos of actions and varying spatial sizes of moving objects. Most of the existing works handle complex spatio-temporal scale variations based on input-level or feature-level pyramid mechanisms, which, however, rely on expensive multistream architectures or explore multiscale spatio-temporal features in a fixed manner. To effectively capture complex scale dynamics of spatio-temporal cues in an efficient way, this article proposes a single-stream architecture (SS-Arch.) with single-input namely, adaptive multi-granularity spatio-temporal network (AMS-Net) to model adaptive multi-granularity (Multi-Gran.) Spatio-temporal cues for video action recognition. To this end, our AMS-Net proposes two core components, namely, competitive progressive temporal modeling (CPTM) block and collaborative spatio-temporal pyramid (CSTP) module. They, respectively, capture fine-grained temporal cues and fuse coarse-level spatio-temporal features in an adaptive manner. It admits that AMS-Net can handle subtle variations in visual tempos and fair-sized spatio-temporal dynamics in a unified architecture. Note that our AMS-Net can be flexibly instantiated based on existing deep convolutional neural networks (CNNs) with the proposed CPTM block and CSTP module. The experiments are conducted on eight video benchmarks, and the results show our AMS-Net establishes state-of-the-art (SOTA) performance on fine-grained action recognition (i.e., Diving48 and FineGym), while performing very competitively on widely used Something-Something and Kinetics.

Real-time instance segmentation of recyclables from highly cluttered construction and demolition waste streams.

Efficient recycling strategies are crucial for mitigating the adverse environmental impacts of escalating construction and demolition waste (CDW). While automated identification via deep-learning is a promising direction, localizing CDW recyclables is uniquely challenging due to significant clutter and compositional complexity. Recognizing that accurate and fast localization is a strong prerequisite for swift robotic action, this study provides a comprehensive assessment of state-of-the-art (s.o.t.a) real-time instance segmentation of recyclables from complex CDW streams. Unlike previous studies that simplify the real-world conditions, this study curates and employs a high-quality CDW instance segmentation dataset tailored to capture often-ignored domain intricacies such as deformation, contamination, intricate class distinctions, scale variations, and high levels of clutter, ensuring practical applicability. The results show that even advanced networks struggle with complexity of real CDW streams due to high clutter, with segmentation accuracy not surpassing 50%. To address this, the study proposes integrating patch-based inferencing techniques, which help the model focus on cluttered regions more effectively, boosting overall performance by a notable 12.9%. Additionally, to enhance the zero-shot capabilities of s.o.t.a prompt-based real-time segmentation for identifying CDW recyclables, a simple yet effective domain-transfer framework is proposed, substantially increasing number of correct mask predictions by 68%. This study serves as a practical reference for applying deep-learning tools in automated environmental management tasks such as waste sorting and suggests the best architectural framework for practical use.

Масштабування українського бізнесу на європейський ринок

Based on the analysis of information sources on business scaling, enterprises were categorized by size and, in turn, by the type of activity of the enterprise. Initially, we analyzed information on Ukraine's cooperation with other countries. The analysis of statistics allows you to identify the countries that show the greatest demand for the products or services that your company produces. This will help you focus your marketing efforts and resources on the most profitable areas.Export statistics indicate the existence of trade barriers, such as duties, quotas, technical regulations, etc. We took into account business scaling strategies for international markets and selected scaling variations by content, in turn, we identified: franchising, branches, subsidiary, outsourcing, joint venture. The next step was to study the strategies of successful Ukrainian companies that have entered foreign markets. In particular, the following brands were identified: “Ruslan Baginsky, M-TAC, and Sandora. The factors that directly contributed to this are analyzed and general patterns and trends are identified. A capsule model of business scaling to the European market has been developed, which provides for the entry of a product into the European market with a 100% profit. This model will allow the company to minimize risks, optimize costs, and quickly adapt to changes in the environment. The European market has its own peculiarities when expanding your business, in particular: each EU country has its own laws and regulations governing business activities. This applies to technical regulations, tax legislation, and labor relations. The EU has a strong consumer protection system. This means that companies must be prepared for various types of claims and lawsuits. The creation of such a model will allow entrepreneurs to scale up virtually risk-free, which in turn will increase foreign exchange earnings and demonstrate a certain level of the country. By entering foreign markets, Ukrainian businesses will take a strategic step that will open up new opportunities for growth and development. Given the current factors, this model may well help entrepreneurs to scale up.

An improved multi-scale feature extraction network for medical image segmentation.

The use of U-Net and its variations has led to significant advancements in medical image segmentation. However, the encoder-decoder structures of these models often lose spatial information during downsampling. Skip connections can help address this issue; however, they may also introduce excessive irrelevant background information. Additionally, medical images display significant scale variations and complex tissue structures, making it challenging for existing models to accurately separate tissues from the background. To address these issues, we developed the Res2Net-ConvFormer-Dilation-UNet (Res2-CD-UNet), a multi-scale feature extraction network for medical image segmentation. This study presents a novel U-shaped segmentation network that employs Res2Net as the backbone and incorporates a convolution-style transformer in the encoding stage to enhance global attention. Additionally, a novel channel feature fusion block (CFFB) has been introduced in the skip connection stage to minimize the effects of background noise. The proposed model was evaluated using publicly available datasets, Synapse and Seg.A.2023. Using the Synapse dataset, the average dice similarity coefficient (DSC) reached 83.92%, which was 1.96% higher than the suboptimal model, and the average Hausdorff distance (HD) was 14.51 mm. Among the eight organs evaluated, optimal results were achieved for four organs. Similarly, using the Seg.A.2023 dataset, the proposed model also achieved the best results with an average DSC of 93.27%. The results of this study indicate that the proposed model can more accurately segment regions of interest and better extract multi-scale features in medical images than existing deep-learning algorithms.

The Environmental Light Characteristics of Forest Under Different Logging Regimes.

Light is a fundamental attribute and key abiotic driver in forest ecosystems. Although the ecological effects of light itself is well studied, capturing the complex parameters that constitute the whole light environment remain an intricate research endeavor. Here, we apply the newly introduced environmental light field (ELF) technique in Kibale National Park, Uganda. We captured whole light scenes with repeat photography and processed it to measure both the spectral composition of light in the red-green-blue range, as well as its variation, or "contrast-span", using the newly introduced International System of Units (SI); "lit". We compare across major and globally common utilized forest types-primary, secondary, and selectively logged areas, as well as a completely cleared area as a control. We find that the ELF system is able to effectively capture key aspects of the local light environment across the range of forest types. The distribution of light intensity and its spectral composition across our study is hardly uniform, with primary forest and a clearing showing two orders of magnitude difference in light. Blue light predominates the sky areas of the clearing, indicating the Rayleigh scattering of sunlight in the atmosphere. In general, radiance decrease with increasing intactness of the forest, and selectively logged and primary forest show the most similar environmental light characteristics. Owing to its ability to capture fine scale variations in light across elevation gradients, their spectral characteristics, as well as their intensities, the ELF system should become a useful tool in better quantifying light in ecology. In particular, we discuss its potential use in restoration ecology.

Mutually-Guided Hierarchical Multi-Modal Feature Learning for Referring Image Segmentation

Referring image segmentation aims to locate and segment the target region based on a given textual expression query. The primary challenge is to understand semantics from visual and textual modalities and achieve alignment and matching. Prior works have attempted to address this challenge by leveraging separately pretrained unimodal models to extract global visual and textual features and perform straightforward fusion to establish cross-modal semantic associations. However, these methods often concentrate solely on the global semantics, disregarding the hierarchical semantics of expression and image and struggling with complex and open real scenarios, thus failing to capture critical cross-modal information. To address these limitations, this article introduces an innovative mutually-guided hierarchical multi-modal feature learning scheme. By leveraging the guidance of global visual features, the model mines hierarchical text features from different stages of the text encoder. Simultaneously, the guidance of global textual features is leveraged to aggregate multi-scale visual features. This mutually guided hierarchical feature learning effectively addresses the semantically inaccurate cause by free-form text and naturally occurring scale variations. Furthermore, a Segment Detail Refinement (SDR) module is designed to enhance the model’s spatial detail awareness through attention mapping of low-level visual features and cross-modal features. To evaluate the effectiveness of the proposed approach, extensive experiments are conducted on three widely used referring image object segmentation datasets. The results demonstrate the superiority of the presented method in accurately locating and segmenting objects in images.