Articles

Between Readiness and Reality: EFL Teachers’ Deep Learning Implementation in Indonesia’s Merdeka Curriculum Amid Remote-Region Constraints

/in /by

This study investigates the readiness of EFL teachers in Toraja, a geographically remote region of South Sulawesi, Indonesia, to implement the Deep Learning approach within Indonesia’s Merdeka Curriculum, and examines the systemic, pedagogical, student-related, and infrastructural challenges they encounter during implementation.A sequential explanatory mixed-methods design was employed, involving six purposively selected junior secondary school EFL teachers. Quantitative data were collected through a validated 20-item questionnaire measuring four readiness dimensions (pedagogical, technological, psychological, and institutional) on a five-point Likert scale, analyzed using descriptive statistics. Qualitative data were gathered through in-depth semi-structured interviews and analyzed using reflexive thematic analysis within Miles and Huberman’s interactive framework. Quantitative results revealed Very High overall teacher readiness (M = 4.28, SD = 0.470), with pedagogical and psychological readiness achieving Very High categorization (M = 4.40 each) and technological and institutional readiness achieving High categorization (M = 4.20 and 4.13 respectively). Four of six teachers (66.7%) were classified as Very High readiness. However, qualitative analysis identified four major challenge themes that systematically constrain implementation: (1) systemic institutional constraints inadequate sporadic professional development, rigid curriculum structures, and heavy administrative burden; (2) pedagogical instructional difficulties severe time constraints, challenges implementing inquiry and reflection phases, and authentic assessment design gaps; (3) student-related barriers uneven readiness, limited EFL vocabulary, passive learning habits, and cultural deference norms; and (4) infrastructure and technological limitations limited shared devices, unstable internet, and forced pedagogical regression reducing deep learning quality by up to 50%. This study reveals a critical readiness-reality gap: teachers demonstrate high internal readiness, yet face substantial external constraints that systematically undermine implementation quality. The findings contribute evidence-based insights to the emerging literature on Deep Learning implementation in under-resourced Indonesian EFL contexts and offer targeted recommendations for teachers, school leaders, district authorities, and national policymakers to achieve sustainable implementation in Toraja and comparable remote regions.​

Deep Learning as a Transformative Pedagogical Model for Critical Thinking Development in Indonesian Vocational English Education

/in /by

The integration of critical thinking in vocational English education is increasingly urgent for 21st-century workforce preparation. However, vocational schools in developing countries like Indonesia struggle to move beyond rote memorization toward reflective learning. This study investigates how deep learning is enacted to develop critical thinking in English classrooms and identifies implementation challenges in Indonesian vocational education. Using a qualitative design, in-depth interviews were conducted with five English teachers at vocational high schools in Tana Toraja, Indonesia. Data were analyzed using Miles and Huberman’s interactive model. Findings reveal that teachers enact deep learning through contextualized materials aligned with students’ vocational fields, higher-order questioning, collaborative activities (project-based learning and discussions), facilitative teaching roles, and supportive classroom climates. These practices foster students’ ability to analyze problems, question information, defend arguments, and transfer critical thinking beyond the classroom. However, implementation faces significant challenges: teacher-level factors (time limitations, conceptual gaps, administrative burden); student-related challenges (mixed abilities, low confidence, unpreparedness for independent learning); institutional barriers (limited technology, assessment complexities); and cultural factors where respect for authority hinders questioning. The study implies that sustainable critical thinking development requires multi-level interventions: context-specific professional development, reduced administrative workload, improved infrastructure, curriculum reforms prioritizing depth over breadth, and culturally responsive pedagogies.

Development of Interactive Lift-The-Flap-Book Media Based on Deep Learning Principles to Enhance Elementary Students’ Reading Comprehension

/in /by

This study aimed to develop and examine the validity and effectiveness of a Lift-The-Flap-Book learning medium based on deep learning principles to improve elementary school students’ reading comprehension skills in historical narrative texts. The study was motivated by students’ difficulties in understanding abstract historical texts due to the dominant use of conventional textbooks and teacher-centered learning methods. This research employed a Research and Development (R&D) approach using the 4D development model, consisting of define, design, develop, and disseminate stages. The subjects were upper-grade elementary students at SD Negeri Gucialit 02, Lumajang Regency. Data were collected through expert validation sheets, reading comprehension tests (pretest and posttest), questionnaires, observations, and interviews. The results indicated that the developed media was highly valid, with validation scores of 98% from media experts, 96% from language experts, and 95% from content experts. The effectiveness test showed a significant improvement in students’ reading comprehension skills, as reflected by N-Gain scores of 0.78 in the small-group trial and 0.77 in the large-group trial, both categorized as high. Therefore, the Lift-The-Flap-Book based on deep learning principles is valid, effective, and feasible for enhancing elementary students’ reading comprehension skills.

The Development of a Deep Learning-Based STEAM Project Module to Enhance Students’ Environmental Literacy through an Eco-Enzyme Initiative

/in /by

The global environmental crisis demands innovative educational approaches to build environmental literacy from an early age. This study aims to develop a deep-learning-based project module integrated with STEAM (Science, Technology, Engineering, Arts, Mathematics) through an eco-enzyme project to improve the environmental literacy of fifth-grade students at MIN 1 Sidoarjo. The research employed the 4D development model (Define, Design, Develop, Disseminate) with a qualitative– quantitative approach. Data were collected through questionnaires, interviews, observations, and pretest–posttest assessments, then analyzed descriptively and statistically (N-Gain). Validation results from experts in content, media, and pedagogy indicated that the module was highly valid (average scores of 4.26, 3.8, and 4.3). Small- and large-scale trials demonstrated that the module was practical (average student response of 3.4) and effective in enhancing environmental literacy, with significant improvements in both cognitive (N-Gain = 0.70) and affective (N-Gain = 0.72) domains. The eco-enzyme project also strengthens the dimensions of the Pancasila Student Profile, particularly creativity, independence, and collaboration. The implications of this study affirm that integrating STEAM and deep learning within a contextual project module can create meaningful learning, foster 21st-century skills, and cultivate students’ ecological awareness. Recommendations include implementing similar modules in elementary schools and developing educational policies that support project-based environmental learning.

Lung Disease Classification Using Transfer Learning on Chest X-ray Images

/in /by

Lung diseases remain a significant global health concern, necessitating the development of rapid and accurate diagnostic methods. While previous research has shown the promise of deep learning models, particularly transfer learning with architectures such as ResNet and VGG, limitations persist in evaluation scope, class imbalance handling, and model interpretability. This study proposes an enhanced deep learning framework for multi-label classification of thoracic diseases using chest X-ray images, addressing these gaps through comprehensive evaluation metrics, advanced data augmentation, and explainable AI (XAI) techniques. The NIH ChestX-ray14 dataset is utilized, with class imbalance mitigated via synthetic minority oversampling and weighted focal loss. Multiple state-of-the-art CNN architectures, including EfficientNet and ResNet variants, are benchmarked using precision, recall, F1 Score, AUC, and accuracy. Moreover, Gradient-weighted Class Activation Mapping (Grad-CAM) is integrated to visualize pathological regions, improving clinical interpretability. The offered framework can perform better in all assessment criteria, achieving an AUC of 0.91 with EfficientNet-B0, and provides interpretable outputs critical for deployment in real-world diagnostic settings. This work advances automated radiological diagnosis by addressing key methodological shortcomings and offers a reliable, explainable solution for lung disease detection.

A Review of AI-powered Diagnosis of Rare Diseases

/in /by

The diagnosis of rare diseases presents significant challenges due to their low prevalence, complex symptomatology, and the scarcity of specialized knowledge. However, advancements in Artificial Intelligence (AI) offer promising solutions to these challenges. This review explores the current state of AI-powered diagnostic tools for rare diseases, focusing on the methodologies, algorithms, and platforms utilized in this emerging field. We examine how AI technologies, such as machine learning, deep learning, and natural language processing, are being integrated into clinical practice to enhance diagnostic accuracy and speed. The research also provides the examples that highlight the successes and limitations of AI in this domain, providing insights into how AI can be harnessed to improve patient outcomes in rare disease diagnosis and management.

Forest Tree AI-SDN Firewall: A Hierarchical Architecture for Adaptive Network Security

/in /by

The rapidly evolving of digital environment prompts advanced network security solutions with essential defend against complex cyber threats. However, network security receives a promising boost from the combination of Software-Defined Networking (SDN) and Artificial Intelligence (AI) because which enables real-time control and intelligent decision-making. Real-time management of network resources through SDN allows flexible control while AI boosts the detection of anomalies in large datasets. In this paper we proposed a Forest Tree AI-SDN Firewall with an innovative hierarchical framework that combines these two powerful technologies to provide adaptive network security solutions with scalable and resilient capabilities. The framework draws its design principles from SDN infrastructure based on three separate layers, Root Layer, Trunk Layer and Canopy Layer. Real-time traffic filtering at the Root Layer uses lightweight edge sensors to achieve 98.2% accuracy while its FPGA-accelerated TLS 1.3 inspection system handles 40 Gbps of data. The Trunk Layer uses reinforcement learning algorithms with a federated SDN control plane to achieve dynamic policy optimization through 12ms response times. The Canopy Layer uses deep learning ensemble technology that combines CNN, LSTM and GNN architectures to detect zero-day threats effectively with 99.4% recall and 92% coverage of encrypted traffic analysis. The system achieves 99.2% threat detection precision during benchmark tests while generating 0.8% incorrect alerts and allowing policy updates at speeds 5.2 times faster than conventional security systems. The proposed system evaluating encrypted information and strengthening adversarial resistance together with cross-domain coordination and achieving 38 Gbps/W energy efficiency.

EcoCycle: A Deep Learning-Based Waste Categorization and Management System for Sustainable Smart Cities

/in /by

Waste management is a critical environmental and economic issue worldwide. Existing waste segregation ac- tivities are inefficient, resulting in high landfill contributions and environmental contamination. In this paper, an artificial intelligence-based waste categorization and management system, EcoCycle, is proposed that utilizes deep learning models like VGG16, ResNet50, and DenseNet121 for automatic classification of waste materials. EcoCycle is equipped with a gamification system based on mobile, a marketplace for recyclables supported by blockchain, and an IoT-based network of intelligent bins for real-time monitoring. Experimental results show 92.36% classification accuracy with DenseNet121, which is improved compared to other implementation results. User survey with 500 users shows a 98% positive effect on user experience and increased awareness about sus- tainability issues. The proposed system contributes significantly towards processes related to circular economies and the goals of smart city initiatives, and it has high global applicability potential for urban waste management systems.

Generative AI in the Categorisation of Paediatric Pneumonia on Chest Radiographs

/in /by

Paediatric pneumonia is a leading cause of morbidity and mortality worldwide, necessitating accurate and timely diagnosis. This study explores the application of Generative AI for categorising paediatric pneumonia using chest radiographs. Leveraging deep learning techniques, including Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), we enhance image quality, generate synthetic training data, and improve model generalizability. The proposed framework integrates AI-driven feature extraction, convolutional neural networks (CNNs), and attention mechanisms to improve diagnostic accuracy. The results demonstrate significant improvements in classification performance compared to traditional methods, with a focus on interpretability and clinical usability.

Adapting Deep-Learning in Early Yam Disease Detection Using Lenet-5 (Adam Optimizer) Convolutional Neural Network Architecture to Improve Productivity and Enhance Farmers Social Habits in the Digital Age

/in /by

The agriculture sector faces significant challenges due to diseases affecting crop yields, particularly in yam cultivation. This study explores the adaptation of deep learning techniques for early detection of yam diseases using a LeNet-5 Convolutional Neural Network (CNN) architecture optimized with the Adam optimizer. The fam sides considered are; Ardokola, Zing and Mutum Biu in Taraba State, Nigeria. By leveraging advanced image processing and machine learning methodologies, we aim to develop an effective diagnostic tool that empowers farmers to identify and manage diseases promptly, ultimately improving productivity. This research not only enhances the technological capabilities of farmers in the digital age but also promotes better agricultural practices, fostering social habits that encourage knowledge sharing and community engagement. The proposed system is tested on a comprehensive dataset of yam leaf images, demonstrating its ability to accurately detect various disease conditions at 17.84%. Results indicate a significant improvement in recognition accuracy, suggesting that the integration of AI-driven solutions can transform disease management approaches in yam farming, contributing to sustainable agricultural practices and improved livelihoods for farmers.