Pabna University of Science & Technology

Traffic sign recognition is a critical aspect of intelligent transportation systems that enhances road safety, traffic management, and driver assistance. Notably, while significant research efforts have been devoted to many prevalent traffic signs, Bangladeshi sign recognition remains relatively underexplored due to a lack of traffic sign knowledge and unawareness of the sector individuals. The scarcity of Bangladeshi traffic sign datasets has exacerbated this limitation, hindering progress. Moreover, most Bangladeshi traffic sign research predominantly relies on a limited number of signs, leading to diminished accuracy and the inability to detect all traffic signs. Additionally, existing systems grapple with suboptimal performance accuracy and heightened computational complexity, further emphasizing the research gap. We propose a robust Bangladeshi traffic sign recognition system to address these formidable challenges using a deep learning-based Convolutional Neural Network (CNN) architecture. The main aim is to create a lightweight model that helps identify traffic signs correctly with less computational effort. We used the effective combination of various deep learning layers in the proposed system. A significant contribution of our work lies in creating a novel Bangladesh Traffic Sign Recognition (BDTSR) dataset, addressing the scarcity of data in this domain. This dataset comprises comprehensive information, including 48 types of traffic signs, and these images were taken from different roads in the country. Our dataset aims to fill a critical gap in Bangladeshi traffic sign research and provides a solid foundation for more extensive and inclusive studies in the field. Through this innovative approach, we aim to contribute significantly to the field of BDTSR, filling the gaps in traffic sign recognition and bolstering the accessibility of traffic signs within the Bangladeshi individual traffic community and beyond. Our evaluation of the newly created BDTSR dataset and the existing two benchmark datasets resulted in high recognition accuracies of 99.34%, 96.97% and 98.82% accuracies for the BDTSR, German Traffic Sign Recognition Benchmark (GTSRB) and the Belgian Traffic Sign Dataset (BTSC) respectively. These results highlight the superiority of our model in the Bangladeshi Traffic Sign Recognition domain, outperforming existing models in terms of accuracy and computational efficiency.

Lead-free double halide perovskites, specifically Cs₂TlAX₆ (A = As, Sb; X = Cl, Br, I), offer key advantages over lead-based versions, including strong optical absorption, high structural and thermal stability, superior carrier mobility, tunable band gaps, non-toxicity, and cost-effectiveness. Their mechanical, electrical, optical, and thermal properties were investigated using DFT with the PBE functional under ambient and hydrostatic pressures. Stability was confirmed through formation enthalpy, tolerance factor, and elastic constants. Pugh’s and Poisson’s ratios suggest these compounds are generally ductile (except Cs₂TlAsBr₆), with pressure-enhanced machinability, reduced friction, and increased plastic strain. Band structure analysis using the GGA-PBE approximation shows the direct band gap semiconducting nature of Cs₂TlAX₆ (A = As, Sb; X = Cl, Br, I), where the band gap values are ranging from 0.957 to 1.697 eV. However, the TB-mBJ functional efficiently moderates the GGA-PBE underestimation, yielding refined band gaps between 1.22 and 2.17 eV—values which are more efficient solar applications. Pressure-induced band gap tuning highlights potential for optoelectronic device applications. Their narrow band gaps and strong absorption make them suitable for solar cells, while high infrared reflectivity and low thermal conductivity indicate potential as thermal barrier coatings (TBCs).

This study investigates the determinants influencing climate change adaptation strategies among rice farmers in the Bogura District of Bangladesh, a region highly vulnerable to climatic variability and agricultural stress. Given the sector’s dependence on rainfall and temperature, understanding the socio-economic and institutional drivers of adaptive behavior is vital for ensuring sustainable rice production and food security. A total of 227 rice farmers were surveyed using semi-structured questionnaires, supplemented by interviews and focus group discussions for contextual validation. Descriptive statistics and Chi-square tests were applied to assess associations between farmers’ socio-economic characteristics and their chosen adaptation strategies. The results were interpreted using odds ratios and 95

Early diagnosis is crucial to preventing irreversible blindness, as retinal disorders are a major cause of vision impairment globally. Manual examination of retinal fundus images is time-consuming, subjective, and prone to diagnostic variability. Although convolutional neural networks (CNNs) have exhibited promising performance in automated retinal illness diagnosis, their limited ability to grasp global contextual dependencies hampers their effectiveness in multi-class classification tasks. This paper provides a robust hybrid model combining CNN and Transformer frameworks with channel attention for the automated classification of multi-class retinal diseases. The proposed method merges a pre-trained EfficientNet backbone for robust local feature extraction with an Efficient Channel Attention mechanism to increase discriminative retinal characteristics. A lightweight Transformer encoder with class token representation is applied to model global contextual relationships among retinal patterns. The model is trained end-to-end utilizing transfer learning, early stopping, adaptive learning rate scheduling, and label smoothing to ensure steady convergence and strong generalization. The system is evaluated on a publicly available retinal fundus picture dataset encompassing seven retinal disease types. Experimental results reveal that the proposed model achieves around 99% accuracy and F1-score, exceeding existing CNN-based techniques. Confusion matrix and ROC analysis confirm significant class separability, whereas precision-recall evaluation indicates balanced and trustworthy diagnostic performance. Attention-based explainability visualizations reveal that the model concentrates on clinically relevant retinal regions, boosting interpretability and trustworthiness. These results indicate that the proposed CNN-Transformer framework provides an accurate, robust, and interpretable solution for automated retinal disease diagnosis and has high promise for deployment in real-world ophthalmic screening and clinical decision-support systems.

Lead-free halide double perovskites K2ScInX6 (X = F, Cl, Br, I) are systematically investigated using firstprinciples DFT to assess their suitability for optoelectronic applications. All compositions crystallize in the cubic Fm-3m phase with increasing lattice constants from F to I. Tolerance and octahedral factors, together with negative formation energies and phonon spectra free of imaginary modes, confirm structural, thermodynamic, and dynamical stability, with K2ScInCl6 exhibiting optimal geometric stability. Electronic calculations reveal direct band gaps at the L-point, tunable via halide substitution. Band gaps range from the visible to near-infrared, refined by SOC and HSE06 treatments. Mechanical analysis satisfies Born stability criteria, showing ductile behavior for Cl-based compounds and higher stiffness for F-based ones. Optical properties demonstrate strong dielectric response, efficient absorption, and increasing refractive index from F to I. Overall, K2ScInX6 emerges as a stable, lead-free, and optically versatile material family for photovoltaic and optoelectronic devices.