Search Results (250)

Class-incremental learning (CIL) in deep neural networks is affected by catastrophic forgetting (CF), where acquiring knowledge of new classes leads to the significant degradation of previously learned representations. This challenge is particularly severe in medical image analysis, where costly, expertise-dependent annotations frequently contain pervasive and hard-to-detect noisy labels that substantially compromise model performance. While existing approaches have predominantly addressed CF and noisy labels as separate problems, their combined effects remain largely unexplored. To address this critical gap, this paper presents a dual-stage clean-sample selection method for Incremental Noisy Label Learning (DSCNL). Our approach comprises two key components: (1) a dual-stage clean-sample selection module that identifies and leverages high-confidence samples to guide the learning of reliable representations while mitigating noise propagation during training, and (2) an experience soft-replay strategy for memory rehearsal to improve the model’s robustness and generalization in the presence of historical noisy labels. This integrated framework effectively suppresses the adverse influence of noisy labels while simultaneously alleviating catastrophic forgetting. Extensive evaluations on public medical image datasets demonstrate that DSCNL consistently outperforms state-of-the-art CIL methods across diverse classification tasks. The proposed method boosts the average accuracy by 55% and 31% compared with baseline methods on datasets with different noise levels, and achieves an average noise reduction rate of 73% under original noise conditions, highlighting its effectiveness and applicability in real-world medical imaging scenarios. Full article

This study proposes a deep metric learning-based pavement distress classification method to address critical limitations in conventional approaches, including their dependency on large training datasets and inability to incrementally learn new categories. To resolve high intra-class variance and low inter-class distinction in distress images, we design a CNN head with multi-cluster centroins trained via SoftTriple loss, simultaneously maximizing inter-class separation while establishing multiple intra-class centers. An adaptive weighting strategy combining sample similarity and class priors mitigates data imbalance, while soft-label techniques reduce labeling noise by evaluating similarity against support-set exemplars. Evaluations on the UAV-PDD2023 dataset demonstrate superior performance—3.2% higher macro-recall than supervised learning, and 6.7%/8.5% improvements in macro-F1/weighted-F1 over iCaRL incremental learning—validating the method’s effectiveness for real-world road inspection scenarios with evolving distress types and limited annotation. Full article

Despite the economic importance of Protected Designation of Origin (PDO) cheeses in Italy, little research has examined how label attributes affect price premiums. For Italian cheese producers, especially those investing in PDO certification, understanding which attributes generate premiums is crucial for sustainable business strategies. This study examined attributes displayed on 420 validated cheese labels collected across Italy in 2022, focusing on hard cheese, fresh soft cheese, and string cheese. A content analysis was conducted to identify and categorize the attributes displayed on cheese labels. Following this, the hedonic pricing method, supported by multiple linear regression analysis, was used to assess the impact of these attributes—along with brand and distribution channel—on product pricing. Our findings reveal that sustainability attributes show particularly strong effects on price premiums. PDO certification generates significant premiums prominently for hard and fresh soft cheeses, cow breed information for string cheese, while specialized retail channels create higher prices for fresh soft and string cheeses. While brand–price relationships are heterogeneous, the study provides evidence of their impact. These insights enable cheese producers, marketers, and retailers to strategically prioritize product attributes, optimize distribution channels, and make informed decisions about brand positioning to maximize value in competitive cheese markets. Full article

High-amylose wheat (HAW), developed through non-genetic modification, addresses the growing demand for clean-label and nutritionally enhanced food products. This study systematically investigated the effects of heat-moisture treatment (HMT; 20% and 25% moisture levels) on the physicochemical properties and cookie-making performance of HAW flour (HAWF) and soft wheat flour (SWF). HMT promoted moisture-induced agglomeration, leading to increased particle size, reduced damaged starch content, and enhanced water and sucrose solvent retention capacities. Although the amylose content remained largely unchanged, pasting behavior was differentially affected, with increased viscosities in SWF and slight decreases in HAWF. Thermal analyses demonstrated elevated gelatinization temperatures, indicating improved thermal stability, while X-ray diffraction revealed alterations in starch crystallinity. Furthermore, HMT weakened gluten strength and modified dough rheology, effects more pronounced in HAWF. Cookies prepared from HMT-treated flours exhibited larger diameters, greater spread ratios, and reduced heights. In vitro digestibility assays showed a marked reduction in rapidly digestible starch and increases in slowly digestible and resistant starch fractions, particularly in HAWF cookies. Collectively, these findings establish HMT as an effective strategy for modulating flour functionality and enhancing cookie quality, while concurrently improving the nutritional profile through the alteration of starch digestibility characteristics. Full article

Bearing fault diagnosis under varying operating conditions faces challenges of domain shift and labeled data scarcity. This paper proposes a dual-stream hybrid-domain adaptation network (DS-HDA Net) that fuses CNN-extracted time-domain features with MLP-processed frequency-domain features for comprehensive fault representation. The method employs hierarchical domain adaptation: marginal distribution adaptation (MDA) for global alignment and conditional domain adaptation (CDA) for class-conditional alignment. A novel soft pseudo-label generation mechanism combining Gaussian mixture models (GMMs) with the Mahalanobis distance provides reliable supervisory signals for unlabeled target domain data. Extensive experiments on the Paderborn University and Jiangnan University datasets demonstrate that DS-HDA Net achieves average accuracy values of 99.43% and 99.56%, respectively, significantly outperforming state-of-the-art methods. The approach effectively addresses bearing fault diagnosis under complex operating conditions with minimal labeled data requirements. Full article

Listeriosis, the disease caused by the bacterium L. monocytogenes, can take invasive forms, with severe complications such as septicemia or meningitis, mainly affecting vulnerable people, such as pregnant women, the elderly, and immunocompromised people. The main transmission is through the consumption of contaminated food, and unpasteurized dairy products are common sources of infection. Due to the high mortality and the difficulty in eliminating the bacterium from the production environment, rigorous hygiene and control measures are essential to prevent the spread of Listeria in the food chain, and research on biofilm formation and bacterial resistance is vital to improve food safety. Dairy products, raw milk, and soft cheeses are among the most vulnerable to contamination with L. monocytogenes, especially due to pH values and low-temperature storage conditions. This paper presents a synthesis of the specialized literature on methods to reduce the incidence of L. monocytogenes in milk and dairy products. Conventional strategies, such as pasteurization and the use of chemical disinfectants, are effective but can affect food quality. Specialists have turned their attention to innovative and safer approaches, such as biocontrol and the use of nonthermal methods, such as pulsed electric fields, irradiation, and nanotechnology. Barrier technology, which combines several methods, has demonstrated superior efficiency in combating the bacterium without compromising product quality. Additionally, lactic acid bacteria (LAB) and bacteriocins are examples of biopreservation techniques that provide a future option while preserving food safety. Natural preservatives, especially those derived from plants and fruits, are promising alternatives to synthetic compounds. Future solutions should focus on developing commercial formulations that optimize these properties and meet consumer demands for healthy, environmentally friendly, and clean-label products. Full article

A composite source model consists of an intrinsic state and an extrinsic observation. The fundamental performance limit of reproducing the intrinsic state is characterized by the indirect rate–distortion function. In a remote classification application, a source encoder encodes the extrinsic observation (e.g., image) into bits, and a source decoder plays the role of a classifier that reproduces the intrinsic state (e.g., label of image). In this work, we characterize the general structure of the optimal transition probability distribution, achieving the indirect rate–distortion function. This optimal solution can be interpreted as a “soft classifier”, which generalizes the conventionally adopted “classify-then-compress” scheme. We then apply the soft classification to aid the lossy compression of the extrinsic observation of a composite source. This leads to a coding scheme that exploits the soft classifier to guide reproduction, outperforming existing coding schemes without classification or with hard classification. Full article

Background: The majority of Australian children exceed the World Health Organization’s recommended dietary intake of free sugar, particularly through the consumption of sugar-sweetened beverages. Front-of-pack nutrition labels increase perceived risk and deter the consumption of sugar-sweetened beverages. However, past studies of young children have focused almost exclusively on a parent’s choice of beverage for children. This study investigated the influence of four label designs (text-based warning, tooth decay pictorial, teaspoons of sugar, and Health Star Rating) on the beverage choices of N = 1229 Australian children (aged 4–11 years) and their parents. Methods: In an online vending machine scenario, parent–child dyads were separately asked to select which beverage they would choose for themselves before and after being randomised to one label condition. The beverages displayed included 100% fruit juice, soft drink, soft drink with a non-nutritive sweetener, flavoured milk, plain milk and bottled water. Beverage healthiness was determined by a 1–10 rating based on a review by a panel of experts (10 dietitians and nutritionists). Results: Mixed-model ANOVAs showed that for parents, each label design performed comparably; however, for children, small but significant differences were seen in the effectiveness of different label designs, with the teaspoons of sugar label, text-based warning, and tooth decay pictorial found to be more impactful in promoting healthier drink choices than the Health Star Rating. Conclusions: These findings can inform public health advocacy efforts to improve food labelling and could be incorporated into educational resources to help children understand the nutritional profiles of different sugary drinks. Full article

The creation of surface relief gratings using hydrogels for label-free biomolecule detection represents a significant advance in the development of versatile, cutting-edge biosensors. Central to this innovation is the formulation of materials with enhanced mechanical properties, especially for applications in soft, wearable technologies. In this work, we have developed novel biofunctional hydrogels that incorporate slide-ring supramolecular structures into their network, enabling the production of surface relief gratings with superior mechanical characteristics for biomolecule detection without the need for labels. These hydrogels, functionalized with bovine serum albumin and goat anti-rabbit antibodies, demonstrated excellent selectivity and sensitivity toward anti-bovine serum albumin and rabbit IgGs in blood serum, evaluated using a label-free format. Remarkably, the new materials matched the analytical performance of conventional hydrogels based on static networks while offering dramatically improved toughness and elasticity, with a compressive modulus comparable to human skin. This demonstrates the potential of slide-ring hydrogels for fabricating robust, label-free biosensing platforms. Furthermore, the flexibility of this system allows for the incorporation of various recognition elements tailored to specific applications. Full article

Excessive sugar consumption has been reported to be associated with various health issues such as obesity, diabetes, cardiovascular diseases, and dental problems. In the UK, effective strategies have been implemented to reduce sugar intake, including the Change4Life Sugar Smart campaign, product reformulation, traffic light labelling, portion control, and the Soft Drinks Industry Levy (SDIL). This review of empirical studies (n = 11) shows that product reformulation, especially in beverages and packaged foods, is effective, as consumers can prefer reduced-sugar alternatives when clearly labelled. The UK traffic light labelling scheme and portion control were also reported to help consumers make informed, healthier food choices. The SDIL, introduced in 2018, was also found to significantly lower sugary beverage consumption. While progress is evident, further nutrition education, public awareness, particularly for people with low socioeconomic status, and more comprehensive policies for long-term positive dietary behavioural shift are essential to limit diseases and conditions associated with high sugar consumption. Future research must evaluate the combined effects of these interventions and examine their long-term effectiveness across diverse population groups. Full article

Soft and flexible capacitive tactile sensors are vital in prosthetics, wearable health monitoring, and soft robotics applications. However, achieving accurate real-time force detection and spatial localization remains a significant challenge, especially in dynamic, non-rigid environments like prosthetic liners. This study presents a real-time force point detection and tracking system using a custom-fabricated soft elastomeric capacitive sensor array in conjunction with image processing and machine learning techniques. The system integrates Otsu’s thresholding, Connected Component Labeling, and a tailored cluster-tracking algorithm for anomaly detection, enabling real-time localization within 1 ms. A $6\times 6$ Dragon Skin-based sensor array was fabricated, embedded with copper yarn electrodes, and evaluated using a UR3e robotic arm and a Schunk force-torque sensor to generate controlled stimuli. The fabricated tactile sensor measures the applied force from 1 to 3 N. Sensor output was captured via a MUCA breakout board and Arduino Nano 33 IoT, transmitting the Ratio of Mutual Capacitance data for further analysis. A Python-based processing pipeline filters and visualizes the data with real-time clustering and adaptive thresholding. Machine learning models such as linear regression, Support Vector Machine, decision tree, and Gaussian Process Regression were evaluated to correlate force with capacitance values. Decision Tree Regression achieved the highest performance ($R^2=0.9996, RMSE=0.0446$), providing an effective correlation factor of $51.76$ for force estimation. The system offers robust performance in complex interactions and a scalable solution for soft robotics and prosthetic force mapping, supporting health monitoring, safe automation, and medical diagnostics. Full article

Due to access to the source data during the transfer phase, conventional domain adaptation works have recently raised safety and privacy concerns. More research attention thus shifts to a more practical setting known as source-data-free domain adaptation (SFDA). The new challenge is how to obtain reliable semantic supervision in the absence of source domain training data and the labels on the target domain. To that end, in this work, we introduce a novel Gradual Geometry-Guided Knowledge Distillation (G2KD) approach for SFDA. Specifically, to address the lack of supervision, we used local geometry of data to construct a more credible probability distribution over the potential categories, termed geometry-guided knowledge. Then, knowledge distillation was adopted to integrate this extra information for boosting the adaptation. More specifically, first, we constructed a neighborhood geometry for any target data using a similarity comparison on the whole target dataset. Second, based on pre-obtained semantic estimation by clustering, we mined soft semantic representations expressing the geometry-guided knowledge by semantic fusion. Third, using the soften labels, we performed knowledge distillation regulated by the new objective. Considering the unsupervised setting of SFDA, in addition to the distillation loss and student loss, we introduced a mixed entropy regulator that minimized the entropy of individual data as well as maximized the mutual entropy with augmentation data to utilize neighbor relation. Our contribution is that, through local geometry discovery with semantic representation and self-knowledge distillation, the semantic information hidden in the local structures is transformed to effective semantic self-supervision. Also, our knowledge distillation works in a gradual way that is helpful to capture the dynamic variations in the local geometry, mitigating the previous guidance degradation and deviation at the same time. Extensive experiments on five challenging benchmarks confirmed the state-of-the-art performance of our method. Full article

Semi-supervised clustering can be viewed as a clustering paradigm that exploits both labeled and unlabeled data to steer learning accurate data clusters and avoid local minimum solutions. Nonetheless, the attempts to refine existing semi-supervised clustering methods are relatively limited when compared to the advancements witnessed in the current benchmark methods in fully unsupervised clustering. This research introduces a novel semi-supervised method for deep clustering that leverages deep neural networks and fuzzy memberships to better capture the data partitions. In particular, the proposed Dual-Constraint-based Semi-Supervised Deep Clustering (DC-SSDEC) method utilizes two sets of pairwise soft constraints; “should-link” and “shouldNot-link”, to guide the clustering process. The intended clustering task is expressed as an optimization of a newly designed objective function. Additionally, DC-SSDEC performance was evaluated through comprehensive experiments using three real-world and benchmark datasets. Moreover, a comparison with related state-of-the-art clustering techniques was conducted to showcase the DC-SSDEC outperformance. In particular, DC-SSDEC significance consists of the proposed dual-constraint formulation and its integration into a novel objective function. This contribution yielded an improvement in the resulting clustering performance compared to relevant state-of-the-art approaches. In addition, the assessment of the proposed model using real-world datasets represents another contribution of this research. In fact, increases of 3.25%, 1.44%, and 1.82% in the clustering accuracy were gained by DC-SSDEC over the best performing single-constraint-based approach, using MNIST, STL-10, and USPS datasets, respectively. Full article

The triboelectric nanogenerator (TENG) is an efficient mechanical energy harvesting device that exhibits excellent performance in the fields of micro-nano energy harvesting and self-powered sensing. In practical application scenarios, it is very important to monitor the speed of rotational machinery in real time. In order to monitor a wider range of rotational speeds, the TENG based on a machine learning algorithm is designed in this paper. The peak power of the TENG reaches a maximum of 6.6 mW and can instantly light up 65 LEDs connected in series. The results show that machine learning can detect speed, greatly improving the speed detection range. The neural network is trained and tested based on the collected electrical signals at different speeds so as to monitor the health of the machine. For the analysis of the collected experimental data, normalization data and a more practical label assignment method of Gaussian soft coding were considered. The study found that after data normalization, the classification prediction accuracy for different speeds is above 0.9, and the prediction results are stable and efficient. Therefore, the machine learning prediction model for speed monitoring proposed by us can be applied to the early warning and monitoring of rotating machinery speed in actual engineering projects. Full article

Background/Objectives: Little is known about preschool-age children’s food concepts among diverse populations. Grounded in the Theory of Mind and Naïve Biology, the primary aim of this study was to describe Black and Hispanic preschool-age children’s food concepts. A secondary aim was to determine the feasibility of collecting data from preschool-age children via a video conferencing platform. Methods: Preliminary qualitative descriptive study. A purposive sample of nine 4- to 6-year-old children ($\overline{\mathrm{x}}$ age = 4.9; Black, n = 7; Hispanic, n = 2), mostly female (n = 7) participated. Children generated two free lists: foods they think of, and foods they eat, reported mouthfeel of 16 foods, and performed a constrained card sort with rationale. Results: All children were able to use the video conference platform. Foods that Black and Hispanic children frequently listed as thought of ($\overline{\mathrm{x}}$ = 6.75) included chicken, rice, carrots, and apples; those frequently listed as foods they eat ($\overline{\mathrm{x}}$ = 8.33) included pancakes and grapes. Black and Hispanic children used various lexicon such as warm, soft, crunchy, and “ouchy” to describe mouthfeel. All preschool-age children sorted foods into piles (range 4–20 piles). Younger children used discrete labels to categorize foods and created many piles while older children used broader labels and created fewer piles. Conclusions: This is the first study to add to the literature about Black and Hispanic preschool-age children’s food concepts before receiving formal education about nutrition. Additionally, we highlight the novel and successful use of ethnographic techniques via internet video conferencing. Subtle differences in their experiential knowledge about food reflect culturally salient qualities that are critical to consider when developing interventions to promote healthy eating behavior. Full article