In resource-constrained educational settings, particularly in developing nations like Ghana, the integration of artificial intelligence (AI) for personalized learning is hindered by unreliable internet connectivity. This paper introduces a novel framework that leverages edge computing to deliver adaptive learning systems offline, addressing the digital divide. The proposed approach combines Bayesian Knowledge Tracing (BKT) on low-cost Raspberry Pi nodes ($55 base units) with multilingual support for Twi, Ewe, and Ga via AfriBERTa language models, and competency-based digital portfolios for continuous assessment. Aligned with Ghana's 2026 National EdTech Strategy and inspired by implementations like Arizona State University's SolarSPELL, the framework operates in an offline-first mode, requiring no hyperscale cloud infrastructure. It structures reform around three pillars: resilient technological infrastructure, teacher-facilitated AI tutoring, and transformative assessment practices. Implementation guidelines, a cost analysis ($1.03 per student per year), and policy recommendations are provided, ensuring alignment with Sustainable Development Goal 4 (Quality Education). Preliminary field validation with Ghanaian educators shows 85% acceptance and projected learning gains of 0.45-0.55 standard deviations. This replicable model advances educational equity through contextually appropriate AI design, bridging policy aspirations with infrastructural realities.
Under the background of the "Belt and Road" initiative, the number of international students from Central Asian and Russian-speaking countries admitted by vocational colleges in Xinjiang has continued to grow, but they face multiple challenges in the process of cross-cultural adaptation. Based on the theory of cross-cultural adaptation and the perspective of educational digital transformation, this study takes 156 international students from Central Asia in three vocational colleges in Xinjiang as research subjects and explores the application scenarios and practical effects of intelligent auxiliary platforms in their adaptation process through a mixed-methods approach. The study constructs an intelligent support platform comprising three major modules and conducts a one-semester intervention experiment using a quasi-experimental design. The results show that the experimental group significantly outperformed the control group in psychological adaptation (F=17.32, p<0.001), sociocultural adaptation (F=21.45, p<0.001), and academic adaptation (F=14.78, p<0.001). Data analysis of platform usage indicates that emergency affairs guidance, cultural cognition construction, and academic norm learning constitute the three core application scenarios, accounting for 41.2%,34.7%, and 24.1% of total usage frequency, respectively. Qualitative research reveals that the platform's real-time response feature significantly alleviates students' adaptation anxiety, while the contextualized learning module enhances their cross-cultural communication skills. This study provides empirical evidence for the application of intelligent technology in cross-border education management and holds significant reference value for promoting the internationalization process of vocational colleges in border regions.
In response to the industrial pain points of low yield of wild strains, high production costs, and single dosage form in clinical application in daptomycin industrial production, and to implement the goal of cultivating the ability to solve complex engineering problems in the undergraduate engineering education accreditation of the pharmaceutical engineering major, this study carried out research on the construction of high-yield daptomycin engineering strains and the process design of a 10 million-dose annual topical gel formulation workshop. Through genetic engineering technology, the pSET152-cumate-atrA recombinant plasmid was constructed and the cumate-induced expression system was established; with batch production as the core mode, formulation screening, process design, material balance, equipment selection, and GMP-compliant workshop layout design were completed, and a comprehensive technical and economic analysis was conducted. The results showed that the recombinant plasmid was successfully constructed and positive clones were verified by colony PCR and plasmid PCR, providing technical support for high-yield industrial strain construction; the optimal prescription and production process of daptomycin topical gel were determined, and the workshop layout meeting D and C grade clean area standards was completed with matched core equipment; the project achieved favorable after-tax annual net profit, cost-profit ratio and economic safety rate, showing good industrial feasibility and economic benefits. This study completed the practical teaching closed loop of the entire chain of "strain modification - formulation research and development - engineering design - industrial implementation" for the pharmaceutical engineering major, which can provide a reference model for undergraduate graduation design teaching and engineering practice ability cultivation of similar majors.
With the accelerated pace of urbanization, urban transportation has become increasingly complex, and residents have an increasingly urgent demand for convenient and efficient bus query services. This study aims to design and implement a bus information management system based on the Spring Boot framework to improve the convenience of obtaining bus travel information. The system adopts the B/S architecture, with the front-end building the user interface using HTML, CSS, JavaScript and the Vue.js framework to achieve friendly interaction with users; the back-end is based on the Spring Boot framework, combined with the MyBatis persistence layer framework to operate the MySQL database for data storage and management. The system has core functions such as line query, station query, and transfer query. Line query supports users to quickly obtain detailed information of the line by entering the line name, including passing stations, first and last bus times, fares, etc.; station query allows users to enter the station name to query all buses passing through the station; transfer query intelligently plans the optimal transfer scheme according to the starting and ending points entered by users. After testing, the system has achieved the expected goals in terms of function realization, performance and user experience, providing strong support for urban residents' bus travel and having high practical value and promotion potential.
In recent years, with the acceleration of urbanization and the continuous increase in the number of vehicles, license plate recognition technology has been widely applied in fields such as intelligent traffic management and security monitoring. However, traditional license plate recognition systems have certain limitations in recognition accuracy and real-time performance, such as false detection and slow recognition speed. To address these issues, an efficient and accurate Chinese license plate recognition system is urgently needed. By combining YOLO v8 and CRNN, this study not only achieves end-to-end license plate detection and recognition but also overcomes the shortcomings of traditional methods, thereby improving recognition accuracy and real-time performance. The main work of this research is to utilize YOLO v8 and CRNN technologies to achieve high accuracy and efficiency of the license plate recognition system. This paper first introduces the basic knowledge of YOLO v8 and CRNN deep learning models and their application advantages in object detection and text recognition. Subsequently, the construction process of the Chinese license plate recognition system combining YOLO v8 and CRNN is elaborated, including basic principles and training steps. In the experimental phase, a comprehensive evaluation and testing of the system are conducted using public license plate datasets and self-made datasets containing license plate images under various complex conditions. Experimental data demonstrate that the system performs excellently in the accuracy and stability of Chinese license plate recognition, with an overall recognition accuracy of 94%. Finally, the application effects of the system and possible directions for future improvement are analyzed, emphasizing its application potential in fields such as intelligent traffic and security monitoring. This study provides a new perspective and solution for the innovation of Chinese license plate recognition technology and holds significant practical value.
As the "navigator" and "calibrator" of educational activities, education evaluation directly determines the direction of educational goals and the quality of talent training. For a long time, education evaluation in China has been dominated by quantitative assessment. Although it has achieved standardization and efficiency in evaluation, it is difficult to take into account the consideration of students' qualitative literacy such as higher-order thinking, essential character, and emotional attitude, falling into the dilemma of "score-only theory". The iterative upgrading of artificial intelligence (AI) technology has provided a historic opportunity to solve this dilemma, promoting the transformation of education evaluation from a quantitative orientation of "digital ruler" to a qualitative empowerment of "situational insight". Based on the practical needs of education evaluation reform, this paper analyzes the limitations of traditional quantitative assessment and the advantages of AI-enabled qualitative evaluation, constructs an AI-driven integrated "quantitative + qualitative" education evaluation system, verifies the implementation path with practical cases and specific charts, and discusses the challenges and solutions in the reform process, providing theoretical reference and practical guidance for education evaluation reform in the new era.
With the rapid advancement of urbanization in China, urban water problems such as frequent waterlogging, water shortage, and water environmental pollution have become increasingly prominent. Sponge city construction has become the core path for urban water system governance and high-quality development of new urbanization. This paper adopts research methods including literature analysis, case study, GIS spatial analysis, and a combination of quantitative and qualitative analysis to systematically sort out the theoretical system and development process of sponge city construction in China, analyze its spatial pattern, implementation effects and existing bottlenecks, and put forward targeted optimization countermeasures. The research results show that China's sponge city construction has gone through four stages and formed a systematic promotion mechanism. The 70 national pilot cities are mainly concentrated in the east of the Hu Huanyong Line. By the end of 2023, the cumulative built-up area of sponge cities nationwide had exceeded 18,000 km², and the number of waterlogging points in pilot cities had been reduced by more than 65% on average. However, the current construction still faces three core bottlenecks: excessive reliance on government finance, imperfect regional adaptive technical standards, and insufficient public participation. Based on this, this paper puts forward optimization paths from four dimensions: multi-scale planning coordination, regional technological innovation, diversified policy guarantee, and universal social participation, to provide theoretical and practical reference for the high-quality development of sponge cities in China, and also provide a research model for practical teaching of geography related majors.
This paper focuses on the task of object detection, utilizing the VOC2007 dataset as the source for training and evaluation, and constructs and implements a Faster R-CNN-based object detection model. As a classic two-stage detector, Faster R-CNN exhibits strong generalization capability and stability. Following the complete object detection pipeline, this experiment sequentially accomplishes four parts: model training, performance evaluation, image inference, and visualization of detection results. During the training phase, a custom VOC data loader and the Faster R-CNN backbone network were employed to effectively learn the features of target categories in the VOC2007 dataset. In the evaluation phase, the VOC standard evaluation protocol was used to compute the Average Precision (AP) for each category and the mean Average Precision (mAP) over the entire validation set. For inference and visualization, the trained model was applied to three test images—test1.jpg, test2.jpg, and test3.jpg—for object detection. The model successfully identified and localized the main target objects in these images, drawing bounding boxes and category labels in distinct colors to fully visualize the detection outcomes. This work establishes a functional object detection system, spanning from data preparation to inference demonstration, and validates the effectiveness of Faster R-CNN in general object detection tasks. Object detection technology holds significant application value in fields such as smart security, autonomous driving, human-computer interaction, and robotic vision. This experiment also lays a foundation for understanding the practical application of deep learning in visual tasks.
With the deep integration and large-scale application of Internet of Things (iot), 5G communication and artificial intelligence technologies, global data is experiencing explosive growth. The limitations of traditional centralized cloud computing architectures in terms of ultra-low latency service demands, massive bandwidth consumption and data privacy protection are becoming increasingly prominent. Edge computing, as an important extension of distributed systems, effectively compensates for the shortcomings of cloud computing in real-time response and bandwidth optimization by sinking computing, storage and application capabilities to the network edge. However, the inherent characteristics of edge nodes, such as limited computing power and heterogeneous resources, make it difficult for them to independently undertake large-scale global optimization tasks. Therefore, "cloud-edge collaboration" has become the core paradigm to solve the contradiction of computing services in the Internet of Things era. From the perspective of distributed system theory, this paper systematically sorts out the evolution context of the cloud-edge collaborative architecture, and deeply analyzes the three-layer logical model of "endpoint device layer - relay edge layer - core cloud computing layer" and its internal collaborative mechanism. Focus on researching key technologies such as computing offloading decision optimization, data consistency maintenance, cloud-edge resource orchestration and communication, and construct multi-objective optimization models and efficient solutions; Conduct empirical analysis based on typical scenarios such as smart cities, Industry 4.0, and intelligent connected vehicles to verify the technical superiority of the cloud-edge collaborative architecture. Finally, the future research directions such as green edge computing, endogenous security mechanisms, and AI adaptive collaboration are prospected. The research results can provide references for the theoretical innovation and engineering implementation of cloud-edge collaborative systems.
2025 marked a pivotal turning point of explosive growth and deep integration of artificial intelligence (AI) technology in the education sector. The role of AI in education is undergoing a systematic shift from traditional "standardized teaching assistance" to "personalized human-machine collaborative enhancement." This article examines the core advancements in AI-driven education during the year, focusing on four key areas: generative AI (GenAI), intelligent teaching agents, edge AI educational hardware, and educational AI governance frameworks. Research indicates that the global educational AI ecosystem demonstrates a trend of balancing technology-driven innovation with educational equity. The study concludes that the current priority has shifted toward the compliant deployment and engineering implementation of AI technologies in educational settings. Future educational paradigms will emphasize the synergy of "AI + Human Intelligence (AI+HI)," propelling AI beyond a mere efficiency tool to evolve into a "human-machine collaborative augmented educational partner" that promotes students' holistic development and enables truly personalized, adaptive teaching.
