Biliran Province State University

Image processing is a technique to translate an image into digital form and execute some operations on it to obtain an improved image or extract some useful information from it. FloraFolium aimed to address the challenges in plant identification, especially in the Philippines, where many plant species are not well-studied or properly recorded. Many people struggle to tell which plants are edible, medicinal, or toxic because of limited access to official guides and reliable information. To solve this problem, the FloraFolium project created a mobile application that uses Convolutional Neural Networks (CNNs) to identify plant leaves and classify them into three categories: edible, medicinal, or toxic. The system was tested and evaluated based on ISO 25010 software quality standards. The results showed high ratings for functionality, usability, and efficiency, making the app reliable for everyday use. While the app performed well, some areas, like security and reliability in unusual conditions, need improvement. The study also found that the image quality greatly affects the system's accuracy. A balanced dataset of 15,000 images was divided into 80% for training and 20% for testing/validation. The model achieved a test accuracy of 99% and an overall validation accuracy of 98.8%, with the best weights saved at epoch 20 during the 30-epoch training period. The FloraFolium app is a helpful tool for outdoor enthusiasts, gardeners, farmers, and anyone interested in learning more about plants. It can also help preserve traditional knowledge about medicinal plants

This study proposes a multidimensional metric for evaluating deep learning models in face image classification, addressing long-standing limitations of accuracy-centered evaluation. While modern architectures such as convolutional neural networks and Vision Transformers demonstrate strong predictive performance, conventional metrics fail to capture deeper structural behaviors, including fairness, representational quality, interpretability, and computational viability. To address this gap, the proposed framework integrates six components: validation F 1 score, linear-probe embedding separability, fairness via skin-tone gap reduction, compute efficiency, anatomical interpretability, and embedding stability, yielding a holistic assessment of model performance. Applied across six benchmark architectures and four Baumann Skin Type tasks, the metric reveals consistent superiority of transformer-based models, which exhibit stronger fairness, stability, and interpretability compared to CNNs. Findings demonstrate that responsible facial model evaluation requires multidimensional criteria that move beyond accuracy to ensure equitable, transparent, and deployment-ready systems. This work contributes toward trustworthy and fairness-aware facial AI.

This cross‑sectional study assessed the prevalence, predictors, and substance use risk profile among 333 incarcerated males at the Freetown Male Correctional Centre. The WHO-ASSIST tool was used for data collection. The risk of substance-related issues was categorised as low, moderate, or high. Logistic regression was done to identify predictors of lifetime substance use (Model I) and recent use within the past three months within or outside prisons (Model II). The prevalence of lifetime substance use was 88.6% (95% CI: 0.85-0.92), recent use 73.9% (95% CI: 0.69-0.79), and polysubstance use 77.2% (95% CI: 0.73-0.82). Tobacco (80.5%), cannabis (70.0%), alcohol (68.2%), and opioids (64.0%) were the most used substances. In adjusted models, older age, unmarried status, and rural residence were significant predictors of both lifetime and recent use, while education status was specifically associated with recent use. Risk profiles revealed a high level of dependence among tobacco users and moderate risk among most cannabis and opioid users. Substance use was notably high, with tobacco, cannabis, and opioids emerging as the primary substances of concern. These findings underscore the burden of substance use and highlight the need for targeted interventions and the integration of cessation programs within the correctional centre.

This study explored the everyday realities of female barangay captains in Leyte, Philippines. It focused on their personal identities and strategies they used to navigate the social and political spheres within their local communities. This study employed a qualitative-phenomenological method. 13 female chiefs served as informants and were purposively selected by the researchers. An interview guide was used as the primary data collection instrument. The analysis follows Braun and Clarke approach to generate themes to describe the informant’s experiences. The results were categorized into experiences, coping with challenges, and aspirations. Seven themes were identified for the experiences: Impactful Service, Personal Fulfillment, Collaborative Environment, Leadership Identity, Personal Hardship, External Adversity, and Leadership Demands. Conversely, three themes were identified as the coping strategies of the female village chiefs, namely Community-Centric Action, Personal Leadership Stance, and Strategic Governance. Moreover, three themes were generated to represent the aspirations: Holistic Community Progress, Enhanced Community Harmony, and Effective Governance. This deep dive into the lifeworld of female village chiefs in the communities in the 3rd congressional district of Leyte has unveiled a myriad of noteworthy leadership stories, struggles, and triumphs, which are evidence of strategic actions and support programs from both the local government and from the community members they led.

Batas Pambansa Blg. 344 has required accessible design in Philippine educational institutions since 1983, and its 2024 Revised Implementing Rules and Regulations now prescribe accessibility dimensions precisely, yet most Philippine schools are reported non-compliant. This study asks why a well-specified standard produces so little measurable compliance in the public basic-education building stock, and examines the framework's architecture rather than its technical content. A qualitative directed content analysis was conducted on seven national legal and administrative instruments, coding each accessibility-relevant provision along three axes — prescriptive specificity, applicability to the existing building stock, and enforcement instrument — within a frame drawn from the implementation and regulatory-compliance literatures. Six structural gaps emerged. The operative duty attaches to the construction license rather than to the building in service, so no instrument creates a verification occasion for a school already in use; retrofit of pre-existing stock is expressed permissively; the penal clause, unchanged since 1983, presupposes a detection that never occurs; the inclusive-education cross-reference reaches only Inclusive Learning Resource Centers; sanitary facilities, the most frequently reported barrier, are the least audited element; and no compliance register or standard audit instrument exists. The framework therefore rests on a negative basis of compliance that the absence of detection has rendered unreachable, while constructing no affirmative basis. Comparison with Malaysia, Singapore and Indonesia indicates that a retrofit duty without verification or finance underperforms, whereas finance can move the existing stock even absent a mandate. The study concludes that the framework is specified but unmonitored. It specifies the funding, auditor calibration, instrument and governance channels through which state university extension could supply third-party audit, addresses the principal objections including unfunded mandates, and proposes four amendments that would institutionalise verification