Stephen Afrifa

Assessing feather condition in broiler breeders is essential for monitoring poultry welfare and identifying feather damage and loss associated with management, health, natural mating, and behavioral challenges. Feather condition can be associated with fertility failures, as featherless areas can suffer scratches from roosters during mating, reducing reproductive success. Manual feather scoring (FS) is labor-intensive, subjective, and difficult to scale in broiler breeder production. This study proposed a stacking ensemble framework for automated multiclass FS classification using RGB images. A total of 67,231 images were categorized into six FS classes (FS_0–FS_5) and used to train three independently trained models: MobileNetV2, EfficientNetB0, and ResNet50. Their class probability outputs were integrated using a Logistic Regression (LR) meta-learner. The proposed stacking ensemble achieved 99.83% accuracy, 99.83% macro precision, 99.86% macro recall, 99.84% macro F1-score, an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.9999, and a quadratic weighted kappa (QWK) of 0.9997. The QWK and ordinal error analysis indicated near-perfect chance-corrected agreement, with no severe cross-category misclassifications. Bird-grouped five-fold cross-validation showed stable performance within the experimental domain evaluated. Gradient-Weighted Class Activation Mapping (Grad-CAM) provided qualitative insight into model attention while also indicating some background activation. The proposed framework shows strong performance on a finer six-class FS task and demonstrates the potential of stacking ensemble learning for objective poultry welfare assessment. Broader cross-farm and cross-device validation remains necessary before practical deployment.

Avian influenza remains a major global threat. It causes recurring outbreaks, economic losses, and public health concerns. Avian influenza virus (AIV) circulates naturally in wild waterfowl. Recent outbreaks have raised concerns about viral evolution and cross-species transmission. Infections in mammals, including dairy cattle and humans, highlight the need for stronger surveillance. This integrative narrative review examines AIV epidemiology, viral evolution, host adaptation, and transmission pathways. It gives particular attention to dust and aerosol exposure. It also considers interactions among wild birds, poultry, and mammals. Structured searches of PubMed and Scopus covered the literature from January 2015 to 12 December 2025. Selected 2026 publications provided additional context. The review synthesizes scientific studies, outbreak investigations, and epidemiological reports. It also examines conventional diagnostics and emerging artificial intelligence (AI), molecular, and sensor-based technologies. Current surveillance and biosecurity approaches face challenges in early detection and coordinated response across production settings. AI-enabled monitoring and advanced diagnostics may improve risk assessment, early warning, and outbreak management. However, these technologies require further field validation. Stronger integration of environmental monitoring, data sharing, and established biosecurity practices is needed. A One Health approach can connect animal, human, and environmental surveillance. Such integration may strengthen AIV preparedness and reduce future outbreak risks.

Reaching, grasping, and placing delicate fresh produce are among the most labor-intensive operations in agricultural production, yet most robotic solutions remain task-specific, single-crop systems that generalize poorly across the morphological diversity of fresh produce. This study investigates whether a large, pre-trained Vision–Language–Action (VLA) model can be adapted with only modest demonstration data to serve as a flexible language-conditioned produce-handling policy on a humanoid robotic platform. The NVIDIA GR00T N1.7-3B foundation model is fine-tuned for a standard 23-degree-of-freedom Unitree G1 humanoid robot fitted with two Dex3-1 three-finger dexterous hands, whose arm and hand joint streams form a 24-dimensional combined manipulation command space. This study also compared the π0.5 VLA model. Demonstrations were collected through a Meta Quest 3 extended-reality teleoperation pipeline augmented with a multi-label recording mode that embeds a natural-language goal in each episode, yielding 605 language-conditioned demonstrations (261,724 synchronized frames) across four produce categories (lemon, peach, pear, and tomato). A dual-camera design paired a wide-angle head camera for global scene context with a wrist-mounted camera for close-range grasp feedback. The results show that the fine-tuned policy achieved a 96.25% individual-trial success rate and 85% on mixed-scene trials requiring retrieval of the language-specified item. However, adapting the π0.5 VLA model to the same data achieved 68.75% individual-trial success. Furthermore, a controlled ablation isolated the role of close-range perception. Without wrist-camera input, the policy reached every target in single-item trials but failed to secure a single grasp, resulting in 0% pick-and-place success despite comparable training loss. Therefore, close-range wrist perception was a precondition for contact in this setting. Together, these results demonstrate that VLA fine-tuning can transform a pre-trained humanoid foundation model into a reliable, data-efficient policy for multi-category fresh produce handling. This work provides a practical step toward flexible humanoid robotic systems for agricultural pick-and-place, sorting, and postharvest automation.

Assessing feather condition in broiler breeders is essential for monitoring poultry welfare and identifying feather damage and loss associated with management, health, natural mating, and behavioral challenges. Feather condition can be associated with fertility failures, as featherless areas can suffer scratches from roosters during mating, reducing reproductive success. Manual feather scoring (FS) is labor-intensive, subjective, and difficult to scale in broiler breeder production. This study proposed an explainable stacking ensemble framework for automated multiclass FS classification using RGB images. A total of 67,231 images were categorized into six FS classes (FS_0-FS_5) and used to train three complementary transfer-learning models: MobileNetV2, EfficientNetB0, and ResNet50. Their class probability outputs were integrated using a Logistic Regression (LR) meta-learner. The proposed stacking ensemble achieved superior performance, with an accuracy of 99.84%, precision of 99.85%, recall of 99.83%, F1-score of 99.84%, AUC of 99.99%, and a quadratic weighted Kappa value of 0.9997, corresponding to 99.97% ordinal agreement. The quadratic weighted kappa demonstrated near-perfect ordinal agreement, confirming minimal severe misclassification across feather-score categories. The experimental results demonstrate high accuracy and reliable inference on workstation hardware, indicating potential for real-time FS monitoring with future optimization for edge-device deployment. Five-fold cross-validation confirmed strong robustness and generalization, while Gradient-weighted Class Activation Mapping (Grad-CAM) visualizations showed that the model focused on biologically meaningful feather regions. Compared with existing state-of-the-art (SOTA) methods, the proposed stacking ensemble achieved higher accuracy on a more challenging six-class task. These findings demonstrate the potential of stacking ensemble learning for objective, scalable, and interpretable poultry welfare monitoring.

In the context of global changes and the transition to a Green Economy, there has been a lack of emphasis in scientific literature to the management of project innovation activities at universities, particularly in collaboration with business and the government. This gap continues despite international experiences that indicate the benefits of industrial symbiosis for regional economic growth, including socioeconomic and environmental issues. The study focuses on the underexplored management of university project innovation activities in collaboration with businesses and the state as the world shifts toward a Green Economy, highlighting the successful use of industrial symbiosis in other countries to boost regional socioeconomic and environmental development. Current engagement channels are insufficient to facilitate the integration of academic ideas into industry throughout the transition to a circular economy. This study intends to establish novel management mechanisms for university innovation activities based on industry symbiosis, as well as optimum collaboration ways for project initiation and implementation. The authors offer successful implementation techniques, such as project formation and realization through partnership with corporations and government bodies, as well as the involvement of bright youngsters. By developing zones of industrial symbiosis and other cooperative ventures, the research hopes to position universities as crucial to building symbiotic chains, eventually boosting university innovation infrastructure efficiency and fostering sustainable regional development. Received: 30 April 2024 | Revised: 19 August 2024 | Accepted: 22 August 2024 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The data generated or analyzed during this study are included in this published article. Author Contribution Statement Daria Y. Mironova: Conceptualization, Methodology, Formal analysis, Writing - original draft, Project administration. Evgeniya A. Pashkova: Validation, Visualization. Alexander G. Budrin: Methodology, Writing - review & editing. Igor V. Baranov: Methodology, Data curation, Writing - review & editing. Vijayakumar Varadarajan: Methodology, Formal analysis, Supervision. Stephen Afrifa: Investigation, Project administration.

Maternal childbirth experiences are crucial indicators of health care quality, patient satisfaction, and emotional health. The increasing use of social media platforms, such as Facebook, provides a unique opportunity to examine public sentiment and narratives around maternity care. However, limited studies have employed deep learning (DL)–based sentiment analysis (SA) to comprehensively analyze childbirth experiences in low‐resource environments. This study utilizes a hybrid technique that integrates unsupervised topic modeling with supervised DL sentiment classification to capture both thematic breadth and emotional tone of birthing experiences. A dataset of Facebook comments was preprocessed and analyzed using word frequency analysis, latent Dirichlet allocation (LDA) for thematic extraction, sentiment classification using convolutional neural networks (CNNs), and robustly optimized BERT pretraining approach (RoBERTa). Statistical correlations between sentiment polarity and hospital service variables were examined using the chi‐square test. The word frequency analysis revealed significant themes such as maternal care, labor and childbirth experiences, informal childbirth discussions, spiritual beliefs, and personal childbirth stories. SA found a majority positive sentiment (5106 instances) with widespread sentiments of trust and joy, although considerable occurrences of fear (1,851), sadness (1,564), and anger (1,114) indicated traumatic delivery experiences. The CNN model outperformed RoBERTa, which had an AUC of 0.9988 and an accuracy of 87%. Statistical study (chi‐square test, p < 0.0001) showed a significant correlation between sentiment polarity and hospital service variables, indicating the influence of treatment quality on patient views. This study emphasizes the significance of SA in assessing maternal health care experiences and improving hospital practices. The findings highlight the need of increased communication, empathetic midwifery care, and patient‐centered approaches in addressing unfavorable childbirth experiences. This study provides policymakers with data‐driven insights to improve maternal health care policies, increase patient experiences, and ensure valuable maternity services.

Climate change has a substantial influence on groundwater levels (GWLs), which are critical for agriculture, safe drinking water, and ecosystem health, which are essential to successful water resource management and adaptation strategies. Recently, there has been an increase in the use of machine learning (ML) and deep learning (DL) models in hydrogeology to estimate GWL in monitoring wells. This study presents a novel technique for predicting GWL changes that uses three independent datasets: historical GWL and climatic variables (CVs) data such as rainfall and temperature influencing groundwater dynamics. In our experimental research, the models’ prediction output on real‐world datasets ensures that the model’s significant patterns are recorded while taking into account the noise in the data, resulting in a perfect balance of bias and variance. The DL models’ results show a significant score of root mean square error (RMSE) between 2.20 and 12.40 and coefficient of determination ( R ‐squared between 0.84–0.99), showing a significant improvement in RMSE and mean absolute error (MAE) in the testing and validation categories, when compared to the current state‐of‐the‐art methods. This study improves our understanding of GWL modeling and provides decision‐makers with a reliable tool for controlling change. The study advances environmental modeling by exhibiting methodological complexity and emphasizes the importance of comprehensive data analysis in water resource management.