Search Results (58)

Colorectal cancer (CRC) is the second most common global malignancy with high mortality, and timely early polyp detection is critical to halt its progression. Yet, polyp image segmentation—an essential tool—faces challenges: blurred edges, small sizes, and artifacts from intestinal folds, bubbles, and mucus. To address these, we proposed a novel segmentation model with multi-scale feature extraction. Its encoder uses Multiscale Attention-based Pyramid Vision Transformer v2 (PVTv2) for hierarchical features (lower-stage modules expand receptive field), while the decoder adopts a Parallel Multi-level Aggregation structure, plus multi-branch and improved reverse attention modules. Ablation experiments validated key modules. Compared to nine state-of-the-art networks across five benchmarks, the model showed superiority: optimal mDice/mIoU on polyp datasets, 0.2% higher mDice than MEGANet on Kvasir-SEG, and outperformance over UHA-Net and CSCA-U-Net on CVC-ClinicDB. Full article

Sitting posture recognition, defined as automatically localizing and categorizing seated human postures, has become essential for large-scale ergonomics assessment and longitudinal health-risk monitoring in classrooms and offices. However, in real-world multi-person scenes, pervasive occlusions and overlaps induce keypoint misalignment, causing global-attention backbones to fail to localize critical local structures. Moreover, annotation scarcity makes small-sample training commonplace, leaving models insufficiently robust to misalignment perturbations and thereby limiting cross-domain generalization. To address these challenges, we propose LAViTSPose, a lightweight cascaded framework for sitting posture recognition. Concretely, a YOLOR-based detector trained with a Range-aware IoU (RaIoU) loss yields tight person crops under partial visibility; ESBody suppresses cross-person leakage and estimates occlusion/head-orientation cues; a compact ViT head (MLiT) with Spatial Displacement Contact (SDC) and a learnable temperature (LT) mechanism performs skeleton-only classification with a local structural-consistency regularizer. From an information-theoretic perspective, our design enhances discriminative feature compactness and reduces structural entropy under occlusion and annotation scarcity. We conducted a systematic evaluation on the USSP dataset, and the results show that LAViTSPose outperforms existing methods on both sitting posture classification and face-orientation recognition while meeting real-time inference requirements. Full article

The rising share of wind generation in power systems, driven by the need to decarbonise the energy sector, is changing the relationship between wind speed and electricity prices. In the case of Poland, this relationship has not been thoroughly investigated, particularly in the aftermath of the restrictive legal changes introduced in 2016, which halted numerous onshore wind investments. Studying this relationship remains necessary to understand the broader market effects of wind speed on electricity prices, especially considering evolving policies and growing interest in renewable energy integration. In this context, this paper analyses wind speed, wind generation, and other relevant datasets in relation to electricity prices using multiple statistical methods, including correlation analysis, regression modelling, and artificial neural networks. The results show that wind speed is a significant factor in setting electricity prices (with a correlation coefficient reaching up to −0.7). The findings indicate that not only is it important to include wind speed as an electricity price indicator, but it is also worth investing in wind generation, since higher wind output can be translated into lower electricity prices. This study contributes to a better understanding of how natural variability in renewable resources translates into electricity market outcomes under policy-constrained conditions. Its innovative aspect lies in combining statistical and machine learning techniques to quantify the influence of wind speed on electricity prices, using updated data from a period of regulatory stagnation. Full article

Electric bicycles in elevators pose serious safety hazards. Fires in the confined space make escape difficult, and recent accidents involving e-bike fires have caused casualties and property damage. To prevent e-bikes from entering elevators and improve public safety, this design employs the Nezha development board as the upper computer for visual detection. It uses deep learning algorithms to recognize hazards like e-bikes. The lower computer orchestrates elevator controls, including voice alarms, door locking, and emergency halt. The system comprises two parts: the upper computer uses the YOLOv11 model for target detection, trained on a custom e-bike image dataset. The lower computer features an elevator control circuit for coordination. The workflow covers target detection algorithm application, dataset creation, and system validation. The experiments show that the YOLOv11 demonstrates superior e-bike detection performance, achieving 96.0% detection accuracy and 92.61% mAP@0.5, outperforming YOLOv3 by 6.77% and YOLOv8 by 15.91% in mAP, significantly outperforming YOLOv3 and YOLOv8. The system accurately identifies e-bikes and triggers safety measures with good practical effectiveness, substantially enhancing elevator safety. Full article

As machine learning becomes more integrated into atmospheric science, XGBoost has gained popularity for its ability to assess the relative contributions of influencing factors in the atmospheric boundary layer height. To examine how these factors vary across seasons, a seasonal analysis is necessary. However, dividing data by season reduces the sample size, which can affect result reliability and complicate factor comparisons. To address these challenges, this study replaces default parameters with grid search optimization and incorporates cross-validation to mitigate dataset limitations. Using XGBoost with four years of data from the atmospheric radiation measurement (ARM) (Southern Great Plains (SGP) C1 site, cross-validation stabilizes correlation coefficient fluctuations from 0.3 to within 0.1. With optimized parameters, the R value can reach 0.81. Analysis of the C1 site reveals that the relative importance of different factors changes across seasons. Lower tropospheric stability (LTS, ~0.53) is the dominant factor at C1 throughout the year. However, during DJF, latent heat flux (LHF, 0.44) surpasses LTS (0.22). In SON, LTS (0.58) becomes more influential than LHF (0.18). Further comparisons among the four long-term SGP sites (C1, E32, E37, and E39) show seasonal variations in relative importance. Notably, during JJA, the differences in the relative importance of the three factors across all sites are lower than in other seasons. This suggests that boundary layer development in the summer is not dominated by a single factor, reflecting a more intricate process likely influenced by seasonal conditions such as enhanced convective activity, higher temperatures, and humidity, which collectively contribute to a balanced distribution of parameter impacts. Furthermore, the relative importance of LTS gradually increases from morning to noon, indicating that LTS becomes more significant as the boundary layer approaches its maximum height. Consequently, the LTS in the early morning in autumn exhibits greater relative importance compared to other seasons. This reflects a faster development of the mixing layer height (MLH) in autumn, suggesting that it is easier to retrieve the MLH from the previous day during this period. The findings enhance understanding of boundary layer evolution and contribute to improved boundary layer parameterization. Full article

The rise of large-scale language models and multimodal transformers has enabled instruction-based policies, such as vision-and-language navigation. To leverage their general world knowledge, we propose multimodal annotations for action options and support selection from a dynamic, describable action space. Our framework employs a multimodal transformer that processes front-facing camera images, light detection and ranging (LIDAR) sensor’s point clouds, and tasks as textual instructions to produce a history-aware decision policy for mobile robot navigation. Our approach leverages a pretrained vision–language encoder and integrates it with a custom causal generative pretrained transformer (GPT) decoder to predict action sequences within a state–action history. We propose a trainable attention score mechanism to efficiently select the most suitable action from a variable set of possible options. Action options are text–image pairs and encoded using the same multimodal encoder employed for environment states. This approach of annotating and dynamically selecting actions is applicable to broader multidomain decision-making tasks. We compared two baseline models, ViLT (vision-and-language transformer) and FLAVA (foundational language and vision alignment), and found that FLAVA achieves superior performance within the constraints of 8 GB video memory usage in the training phase. Experiments were conducted in both simulated and real-world environments using our custom datasets for instructed task completion episodes, demonstrating strong prediction accuracy. These results highlight the potential of multimodal, dynamic action spaces for instruction-based robot navigation and beyond. Full article

Background: Oncogene-induced senescence (OIS) is a tumor-suppressive mechanism that halts uncontrolled cell proliferation in premalignant lesions. Further investigation into its role in colorectal tumorigenesis is essential. We investigated the expression of OIS transcriptomic landscapes in premalignant colorectal adenomas and whether their resolution is part to adenoma-to-carcinoma progression. Methods: Using a publicly available gene expression dataset (GSE117606), we analyzed 66 paired (matched) adenoma–adenocarcinoma samples. Single-sample gene set enrichment analysis (ssGSEA) was performed to assess OIS and senescence-associated secretory phenotype (SASP) signatures, and differential gene expression analysis was conducted to examine key senescence-related genes. Results: OIS and SASP signatures were significantly enriched in adenomas compared to adenocarcinomas (p < 0.05). Pairwise comparisons confirmed that 65% of patients exhibited higher OIS scores in adenomas, while SASP enrichment declined in 59–61% of cases. Several senescence regulators (CDKN1A, CDKN2B, and E2F3), ECM remodeling genes (MMP10 and TIMP2), and NF-κB-driven SASP factors (CCL2, CXCL2, NFKB1, and NFKB2) were significantly downregulated in adenocarcinomas, indicating the resolution of senescence-associated inflammatory signaling during tumor progression. Conclusions: These findings support the predominance of OIS phenotypes in colorectal adenomas, suggesting their potential role as a temporary barrier to tumorigenesis in colorectal cancer. Full article

With the explosive growth of sequencing data, rapidly and accurately classifying and identifying species has become a critical challenge in amplicon analysis research. The internal transcribed spacer (ITS) region is widely used for fungal species classification and identification. However, most existing ITS databases cover limited fungal species diversity, and current classification methods struggle to efficiently handle such large-scale data. This study integrates multiple publicly available databases to construct an ITS sequence database encompassing 93,975 fungal species, making it a resource with broader species diversity for fungal taxonomy. In this study, a fungal classification model named FungiLT is proposed, integrating Transformer and BiLSTM architectures while incorporating a dual-channel feature fusion mechanism. On a dataset where each fungal species is represented by 100 ITS sequences, it achieves a species-level classification accuracy of 98.77%. Compared to BLAST, QIIME2, and the deep learning model CNN_FunBar, FungiLT demonstrates significant advantages in ITS species classification. This study provides a more efficient and accurate solution for large-scale fungal classification tasks and offers new technical support and insights for species annotation in amplicon analysis research. Full article

In long-tail scenarios, models have a very high demand for high-quality data. Information augmentation, as an important class of data-centric methods, has been proposed to improve model performance by expanding the richness and quantity of samples in tail classes. However, the underlying mechanisms behind the effectiveness of information augmentation methods remain underexplored. This has led to reliance on empirical and intricate fine-tuning in the use of information augmentation for long-tail recognition tasks. In this work, we simultaneously consider the richness gain and distribution shift introduced by information augmentation methods and propose effective information gain (EIG) to explore the mechanisms behind the effectiveness of these methods. We find that when the value of the effective information gain appropriately balances the richness gain and distribution shift, the performance of information augmentation methods is fully realized. Comprehensive experiments on long-tail benchmark datasets CIFAR-10-LT, CIFAR-100-LT, and ImageNet-LT demonstrate that using effective information gain to filter augmented data can further enhance model performance without any modifications to the model’s architecture. Therefore, in addition to proposing new model architectures, data-centric approaches also hold significant potential in the field of long-tail recognition. Full article

The problems of missed segmentation with fuzzy boundaries of segmented regions and small regions are common in segmentation tasks, and greatly decrease the accuracy of clinicians’ diagnosis. For this, a new network based on PSPNet, using a coordinate attention mechanism and a mixed loss function for segmentation (CAML-PSPNet), is proposed. Firstly, the coordinate attention module splits the input feature map into horizontal and vertical directions to locate the edge position of the segmentation target. Then, a Mixed Loss function (MLF) is introduced in the model training stage to solve the problem of the low accuracy of small-target tumor segmentation. Finally, the lightweight MobilenetV2 is utilized in backbone feature extraction, which largely reduces the model’s parameter count and enhances computation speed. Three datasets—PrivateLT, Kvasir-SEG and ISIC 2017—are selected for the experimental part, and the experimental results demonstrate significant enhancements in both visual effects and evaluation metrics for the segmentation achieved by CAML-PSPNet. Compared with Deeplabv3, HrNet, U-Net and PSPNet networks, the average intersection rates of CAML-PSPNet are increased by 2.84%, 3.1%, 5.4% and 3.08% on lung cancer data, 7.54%, 3.1%, 5.91% and 8.78% on Kvasir-SEG data, and 1.97%, 0.71%, 3.83% and 0.78% on ISIC 2017 data, respectively. When compared to other methods, CAML-PSPNet has the greatest similarity with the gold standard in boundary segmentation, and effectively enhances the segmentation accuracy for small targets. Full article

The demand for flexible automation in manufacturing has increased, incorporating vision-guided systems for object grasping. However, a key challenge is in-hand error, where discrepancies between the actual and estimated positions of an object in the robot’s gripper impact not only the grasp but also subsequent assembly stages. Corrective strategies used to compensate for misalignment can increase cycle times or rely on pre-labeled datasets, offline training, and validation processes, delaying deployment and limiting adaptability in dynamic industrial environments. Our main contribution is an online self-supervised learning method that automates data collection, training, and evaluation in real time, eliminating the need for offline processes. Building on this, our system collects real-time data during each assembly cycle, using corrective strategies to adjust the data and autonomously labeling them via a self-supervised approach. It then builds and evaluates multiple regression models through an auto machine learning implementation. The system selects the best-performing model to correct the misalignment and dynamically chooses between corrective strategies and the learned model, optimizing the cycle times and improving the performance during the cycle, without halting the production process. Our experiments show a significant reduction in the cycle time while maintaining the performance. Full article

Existing Transformers for 3D human pose and shape estimation models often struggle with computational complexity, particularly when handling high-resolution feature maps. These challenges limit their ability to efficiently utilize fine-grained features, leading to suboptimal performance in accurate body reconstruction. In this work, we propose TransSMPL, a novel Transformer framework built upon the SMPL model, specifically designed to address the challenges of computational complexity and inefficient utilization of high-resolution feature maps in 3D human pose and shape estimation. By replacing HRNet with MobileNetV3 for lightweight feature extraction, applying pruning and quantization techniques, and incorporating an early exit mechanism, TransSMPL significantly reduces both computational cost and memory usage. TransSMPL introduces two key innovations: (1) a multi-scale attention mechanism, reduced from four scales to two, allowing for more efficient global and local feature integration, and (2) a confidence-based early exit strategy, which enables the model to halt further computations when high-confidence predictions are achieved, further enhancing efficiency. Extensive pruning and dynamic quantization are also applied to reduce the model size while maintaining competitive performance. Quantitative and qualitative experiments on the Human3.6M dataset demonstrate the efficacy of TransSMPL. Our model achieves an MPJPE (Mean Per Joint Position Error) of 48.5 mm, reducing the model size by over 16% compared to existing methods while maintaining a similar level of accuracy. Full article

Although imbalanced data have been studied for many years, the problem of data imbalance is still a major problem in the development of machine learning and artificial intelligence. The development of deep learning and artificial intelligence has further expanded the impact of imbalanced data, so studying imbalanced data classification is of practical significance. We propose an image oversampling algorithm based on the influence function and sample weights. Our scheme not only synthesizes high-quality minority class samples but also preserves the original features and information of minority class images. To address the lack of visually reasonable features in SMOTE when synthesizing images, we improve the pre-training model by removing the pooling layer and the fully connected layer in the model, extracting the important features of the image by convolving the image, executing SMOTE interpolation operation on the extracted important features to derive the synthesized image features, and inputting the features into a DCGAN network generator, which maps these features into the high-dimensional image space to generate a realistic image. To verify that our scheme can synthesize high-quality images and thus improve classification accuracy, we conduct experiments on the processed CIFAR10, CIFAR100, and ImageNet-LT datasets. Full article

Brazil’s Forest Code (FC) is a landmark law that, despite dating back to the 1930s, has low compliance. Illegal deforestation continues, and millions of hectares that were set to be reforested remain degraded. Although sector agreements such as the Amazon Soy Moratorium (ASM) have been important in the fight against deforestation, the implementation of the FC represents the key long-term strategy to halt deforestation in the soy supply chain. Here, we used datasets of the boundaries of rural properties, deforestation permits, environmental licensing, and land cover in Mato Grosso to quantify illegal deforestation and analyzed compliance with the Forest Code (FC) on soy farms to explore how loopholes in the implementation of the FC allow deforestation to continue unabated. Our analyses show that between August 2009 and July 2019, soy farms in Mato Grosso State, the largest Brazilian soy producer, were responsible for 15% (or 468.1 thousand hectares) of all land cleared in registered properties. Half of this deforestation was illegal. The FC implementation within these properties has been slow: only 11% of registered soy farms have made it to the final stage of the registration process, thus being considered fully compliant. This novel analysis reinforces that accelerating the implementation of the FC could significantly reduce deforestation and advance the restoration of illegally cleared land particularly in the Cerrado, where 50% of the original cover has already been lost, as well as in the Amazon. By achieving full compliance in the soy sector, Brazil’s position in the international market would be strengthened as a supplier of sustainably produced, deforestation-free commodities. Full article

The TEST pervasive use of social media has highlighted the importance of developing sophisticated models for early information warning systems within online communities. Despite the advancements that have been made, existing models often fail to adequately consider the pivotal role of network topology and temporal dynamics in information dissemination. This results in suboptimal predictions of content propagation patterns. This study introduces the User Propagation Influence-based Linear Threshold (UPI-LT) model, which represents a novel approach to the simulation of information spread. The UPI-LT model introduces an innovative approach to consider the number of active neighboring nodes, incorporating a time decay factor to account for the evolving influence of information over time. The model’s technical innovations include the incorporation of a homophily ratio, which assesses the similarity between users, and a dynamic adjustment of activation thresholds, which reflect a deeper understanding of social influence mechanisms. Empirical results on real-world datasets validate the UPI-LT model’s enhanced predictive capabilities for information spread. Full article