Computer vision is one of the artificial intelligence’s most challenging fields, enabling computers to interpret, analyse and derive meaningful information from the visual world. There are various utilizations of computer vision algorithms, and most of them, from simpler to more complicated, have an object and shape recognition in common. Traditional pen and paper tests are designed in a pre-established format and consist of numerous basic shapes, which designate the important parts of the test itself. With that in mind, many computer vision applications regarding pen and paper tests arise as an opportunity. Massive courses and large schooling organizations mostly conduct their exams in paper format and assess them manually, which imposes a significant burden on the teaching staff. Any kind of automatization that will facilitate the grading process is highly desirable. Hence, an automated answer recognition system in assessment was developed to mitigate the problems above. The system uses images of scanned test pages obtained from the test scanning process and performs the necessary image manipulation steps to increase target recognition accuracy. Further, it manages to identify regions of interest containing multiple-choice questions and contours. Finally, the system verifies obtained results using the knowledge of the whereabouts of the test template regions of interest.

The advent of digital technology has revolutionized numerous aspects of modern life, including the field of assessment and testing. However, paper tests, despite their seemingly archaic nature, continue to hold a prominent position in various assessment domains. The accessibility, familiarity, security, cost-effectiveness, and versatility of paper tests collectively contribute to their continued prominence. Hence, numerous educational institutions responsible for conducting examinations involving a substantial number of candidates continue to rely on paper tests. Consequently, there arises a demand for the possibility of automated assessment of these tests, aiming to alleviate the burden on teaching staff, enhance objectivity in evaluation, and expedite the delivery of test results. Therefore, diverse software systems have been developed, showcasing the capability to automatically score specific question types. Thus, it becomes imperative to categorize related question types systematically, thereby facilitating a preliminary classification based on the content and format of the questions. This classification serves the purpose of enabling effective comparison among existing software solutions. In this research paper, we present the implementation of such a software system using artificial intelligence techniques, progressively expanding its capabilities to evaluate increasingly complex question types, with the ultimate objective of achieving a comprehensive evaluation of all question types encountered in paper-based tests. The system detailed above demonstrated a recognition success rate of 99.89% on a curated dataset consisting of 734,825 multiple-choice answers. For the matching type, it achieved a recognition success rate of 99.91% on 86,450 answers. In the case of short answer type, the system achieved a recognition success rate of 95.40% on 129,675 answers.

In this paper, we describe the possibility of using large language models (LLMs) in education through the development of a modern web system. A web-based platform employing LLMs to produce questions and answers spanning diverse domains, drawing upon Serbian Wikipedia as its primary knowledge repository. Leveraging LLMs for question-and-answer generation represents a novel methodology in quiz development. Additionally, incorporating Wikipedia enhances the accuracy and pertinence of the generated questions, thereby enriching the overall user experience.

Industrial Wireless Sensor Networks (IWSNs) play a vital role in today’s Industrial Internet of Things (IoT) systems. The existing Enhanced Energy Optimization Model has made progress in improving energy use; however, it still faces issues with adapting to changing network traffic, slow clustering convergence, and limited optimization capabilities. To address these problems, this paper presents a hybrid intelligent framework that improves energy savings, clustering convergence, and adaptive network optimization in IWSNs. The new model combines Enhanced Aphid–Ant Mutualism Optimization (EAAM) for efficient clustering, an Improved Artificial Bee Colony (I-ABC) algorithm for ongoing energy parameter adjustments, and a Graph Neural Network (GNN)-based predictor for adapting to changing topologies and traffic conditions. EAAM provides fair cluster-head selection and minimizes control overhead, while I-ABC adjusts transmission power and duty cycles to reduce energy loss. The GNN module captures spatial correlations between sensor nodes to provide proactive energy prediction and adaptive reconfiguration, with lower training overhead than deep reinforcement learning. Experimental results on benchmark IWSN datasets verify that the proposed EAAM–I-ABC–GNN framework exhibits 18%–25% better overall energy efficiency, a 22% increase in network lifetime, and 15% improved clustering stability compared to the baseline EEOM and traditional hybrid methods. These findings validate that the proposed hybrid model substantially improves scalability, flexibility, and sustainability, providing an efficient and intelligent solution for future Industrial IoT networks.

Automatic license-plate recognition (ALPR) systems typically rely on a two-stage pipeline that combines object detection with optical character recognition (OCR), whereas recent vision–language models (VLMs) enable direct image-level recognition without task-specific training. This paper presents a comprehensive comparison between these paradigms under limited target-domain data availability. We evaluate a plate-specific YOLO11x detector combined with four OCR backends and five locally hosted open-weight VLMs, without training or fine-tuning on a study-specific Serbian dataset containing 435 images and 643 visible license plates, annotated with image-level plate counts, registration transcriptions, and plate-level bounding boxes. Performance is assessed using strict end-to-end success, exact-match accuracy, character error rate, readable-plate recall, plate-count accuracy, and inference speed. The comparison is further validated on the OpenALPR-EU and UC3M-LP benchmarks. Finally, we investigate the effect of increasing target-domain annotation budgets by incrementally fine-tuning the YOLO detector on UC3M-LP to determine when a conventional detector–OCR pipeline overtakes zero-shot VLMs. The results demonstrate complementary strengths rather than a universal winner. YOLO11x combined with FastPlateOCR achieves the highest transcription accuracy and is approximately 13× faster than the best-performing VLM; the three general-purpose OCR backends collapse on identical detections, so these figures characterize the strongest available plate-specific recognizer rather than detector–OCR pipelines in general, while Qwen3-VL provides superior image-level plate-count accuracy in the zero-shot setting, an advantage concentrated in single-plate scenes. The annotation-budget study shows that a few hundred labeled localization annotations substantially improve the conventional pipeline, enabling it to surpass zero-shot VLMs in character recovery, although VLMs retain an advantage in image-level plate counting. These findings provide practical guidance for selecting ALPR systems according to annotation availability, deployment constraints, and application objectives.