This review focused on automated detection and classification of wood surface defects, such as knots, cracks, and resin pockets. Machine learning and deep learning methods are increasingly replacing traditional inspection techniques, such as manual checks and imaging tools like X-rays and ultrasound. However, these conventional approaches suffer from significant limitations in accuracy and efficiency. We reviewed both single-stage models, such as you only look once (YOLO) and its variants, and two-stage models like faster region-based convolutional neural networks (R-CNN), and examined their strengths, limitations, and relevance in real industrial applications. We further investigated emerging zero-shot learning approaches that use vision-language models and NLP techniques to detect previously unseen wood defects. Zero-shot models do not rely on annotated training data. These methods offer scalable and flexible solutions, especially in scenarios where collecting labelled samples for all defect types is impractical. The paper also discussed publicly available labelled datasets used to train and test these models, and discussed standard performance metrics such as precision, recall, and mean average precision. This review aimed to support further research and practical improvements in automated wood surface quality assessment by analyzing defect types, detection methods, and evaluation techniques.

This study proposes an algorithm based on an improved YOLOv8s for nighttime vehicle recognition under low light and complex background conditions, addressing issues such as shallow feature loss, low recognition accuracy for occluded targets, and bounding box distortion. The core optimization strategies are as follows: First, RepGFPN (Re-parameterized Generalized Feature Pyramid Network) employs re-parameterized fusion, utilizing a customized fusion module to achieve multi-scale feature interaction. This preserves shallow-level details while eliminating redundant sampling and adjusting channels to enhance fusion efficiency. Second, the DAT (Deformable Attention Transformer) attention mechanism generates reference grids, while the Lightweight Network predicts deformation point offsets. Bilinear interpolation extracts key features, enhancing learning capabilities for small-sized and occluded targets. Third, the MPDIoU (Minimum Point Distance Intersection over Union) loss function minimizes Euclidean distances between bounding box corners, integrating intersection-over-union optimization to mitigate deformation caused by vehicle overlap. Experimental results demonstrate superior performance over the original model: accuracy improved by 6.4%, recall increased by 4.9%, latency reduced by 6.3 ms, mAP0.5 was boosted by 4.4%, and N_Params decreased by 2.5.

Crash diagrams are essential tools in transportation safety analysis, yet their manual preparation remains time-consuming and prone to human variability. This study investigates the use of Vision-Language Models (VLMs) to automate crash diagram generation from police crash reports, focusing on multilane roundabouts as a challenging test case. A three-part structured prompt framework was developed to guide model reasoning through interpretation, extraction, and visual synthesis, while a 10-metric evaluation system was designed to assess diagram quality in terms of semantic accuracy, spatial fidelity, and visual clarity. Three popular models, including GPT-4o, Gemini-1.5-Flash, and Janus-4o, were tested on 79 crash reports. GPT-4o achieved the highest average performance (6.29 out of 10), followed by Gemini-1.5-Flash (5.28) and Janus-4o (3.64). The analysis revealed GPT-4o’s superior spatial reasoning and alignment between extracted and visualized crash data. These results highlight both the promise and current limitations of VLMs in engineering visualization tasks. The study lays the groundwork for integrating generative AI into crash analysis workflows to improve efficiency, consistency, and interpretability.

Large language models (LLMs) are designed to read, reason, and generate natural language, enabling improved access to information, problem solving, and communication. However, their performance depends strongly on prompt design and model scale. Lightweight LLMs (fewer than 5B parameters) often struggle to summarize complex technical documents such as research publications due to domain-specific terminology, dense citations, and mathematical notation, which can increase hallucinations and reduce reliability. We have introduced a hybrid framework that combines Natural Language Processing (NLP) driven preprocessing techniques, sparse input activation, and prompt engineering to enhance the summarization capacity of small-scale models by minimizing hallucinations and maximizing factual accuracy. The pipeline starts with cleaning and segmenting full articles into sections, removing references, citations, and formulas. On this normalized output, salient sentence extraction and keyphrase extraction are performed as part of the sparse input activation module. The activated input and optimized prompt are fed to LLMs of different scales and the summaries are benchmarked with respect to different prompting strategies and model sizes. This workflow significantly enhances the factual accuracy and reliability of summaries generated by lightweight LLMs, making them competitive for complex scientific and technical summarization tasks. Our adaptive prompting and sparse input activation approach significantly boosted summarization quality in lightweight LLMs, improving average ROUGE-1 recall from about 32% to 46% (44% relative gain), human evaluation scores from 66% to 91% (39% relative gain), and coverage ratio from 39% to 60% (53% relative gain) over the baseline.

Optimizing health care services can be considered as an optimization problem, but what exactly should we optimize? We present the problem as a two-objective optimization problem, minimizing cost and maximizing quality. The challenge is that, even if we had concrete quality indicators for all health care sectors, it is not clear how we would solve it. In this paper, we discuss alternative scenarios for solving the problem and the pitfalls we should avoid.

Dementia encompasses a group of syndromes that impair cognitive functions such as memory, reasoning, and the ability to perform daily activities. As populations globally age, nearly 10 million new dementia cases occur annually. Clinical diagnosis remains challenging because symptoms overlap with other conditions and require comprehensive cognitive assessment, highlighting the need for feasible and accurate detection methods. Recent advances in machine learning have highlighted spontaneous speech as a promising noninvasive, cost-effective, and scalable biomarker for dementia detection. In this study, spontaneous speech recordings from the ADReSS dataset and the extended Pitt Corpus were analyzed, consisting of picture description tasks performed by cognitively healthy individuals and participants with Alzheimer’s disease or dementia. Unlike many prior approaches relying on speech-active segments, acoustic features were extracted from entire recordings with the openSMILE toolkit. This recording-level representation reduces the number of feature vectors and provides a computationally efficient framework for dementia classification, while indirectly incorporating pause- and hesitation-related information. Classification models with classifier-based wrapper feature selection were employed to estimate feature importance and identify diagnostically relevant acoustic characteristics. Among the evaluated classifiers, the extreme minimal learning machine emerged as the most computationally efficient method, providing competitive classification accuracy with substantially lower training time in repeated leave-one-subject-out validation. The results demonstrated that the proposed framework is computationally efficient, interpretable, and well-suited as a supportive tool for speech-based dementia assessment.