University of Pitești

Wildlife conservation efforts increasingly depend on automated species classification for processing large-scale camera trap data, yet existing approaches struggle with accuracy and computational efficiency in resource-constrained environments. This paper introduces HARVEST (Hierarchical Attention for Robust Vision Enhancement with Shifted Tokenization), a novel hybrid architecture integrating YOLOv8 object detection with transformer-based classification. The architecture incorporates three key innovations: Shifted Patch Tokenization (SPT) for boundary information preservation, Local Information Enhancer (LIFE) for spatial feature extraction, and Locality-Enhanced Attention (LEA) for adaptive feature integration. The model is evaluated on two comprehensive datasets: a challenging 45-species Ohio State University (OSU) Small Animals dataset exhibiting an extreme class imbalance (6320:1 ratio) and a balanced 6-species African wildlife dataset. The HARVEST demonstrates excellent performance and achieves 85.27% accuracy on the OSU dataset and 94.74% accuracy on the Wildlife dataset with only 13.0M parameters, representing an 85% reduction compared to standard Vision Transformers while maintaining superior performance. The OSU evaluation demonstrates robust performance across highly imbalanced real-world conditions with species sample sizes ranging from 1 to 6320 images, validating practical applicability for conservation scenarios. Qualitative analysis reveals biologically meaningful attention patterns focusing on taxonomically relevant features. The efficient architecture enables real-world deployment in conservation applications, providing a practical solution for automated wildlife monitoring and biodiversity surveillance.

Background and purpose The problem of injury prevention in sports with changing circumstances and, accordingly, the preparation of future physical education and sports specialists for injury prevention in their students is of great relevance both for sports achievements and for maintaining the health of athletes. Purpose: based on a systematic analysis of the literature, to develop and substantiate the concept of an individual approach to injury prevention and to determine ways to train physical education and sports specialists to implement this concept in practical activities in the training process in sports with changing circumstances. Material and methods The search and selection of publications for analysis complied with the 2020 recommendations to ensure transparency and reproducibility. The following scientific electronic databases were selected to search for publications: Scopus, Web of Science, PubMed/MEDLINE and Google Scholar until November 25, 2025. Results The most effective preventive measures were identified: Neuromuscular training (NMT), Strengthening the cortex and stabilizers, Dosing the load, Improving equipment and the environment. Training future specialists in physical education and sports for injury prevention should be conducted in two main areas: 1 - mastering the skills and abilities to prevent injuries in their own sports activities; 2 - mastering the knowledge of injury prevention in sports with changing circumstances and the ability to transfer this knowledge to future students. Conclusions The injury prevention system involves the integration of the following vectors of work and training of specialists: theoretical and methodological, analytical, practical. The theoretical and methodological vector involves the formation of a concept of the individual characteristics of the organism. The analytical vector consists of biomechanical analysis of movement techniques, registration of psychophysiological functions, anthropometric indicators, etc. The practical vector involves monitoring the athlete's condition, applying injury prevention measures, and planning the process of training athletes in sports with changing circumstances, taking into account their individual characteristics.

Warm-up is widely recognized as a fundamental component of athletic preparation and is commonly used to enhance performance and reduce injury risk. Various warm-up strategies have been investigated in recent years, including dynamic warm-up, static stretching, Post-Activation Performance Enhancement (PAPE), foam rolling, and plyometric exercises. However, the relative effectiveness of these approaches remains unclear. The aim of this narrative review was to synthesize current evidence regarding the effects of different warm-up strategies on athletic performance and injury prevention. A literature search was conducted using PubMed, Scopus, Web of Science, and Google Scholar databases, and studies published between 2015 and 2025 were considered. Twenty-two studies met the inclusion criteria. The findings indicated that dynamic warm-up produced the most consistent improvements in sprint performance, agility, jumping ability, and sport-specific skills. PAPE protocols were also associated with enhanced power output and explosive performance. Static stretching primarily improved flexibility and range of motion, whereas foam rolling showed limited effects on performance despite increasing mobility. Plyometric warmup strategies yielded mixed results. In addition, structured warm-up programs incorporating balance, strength, and neuromuscular exercises contributed to injury prevention and improved postural control. Overall, the available evidence suggests that dynamic warm-up and PAPE protocols currently represent the most effective evidence-based strategies for enhancing athletic performance, while structured warm-up programs play an important role in injury prevention.

In recent decades, the rising challenges posed by climate change have prompted investors to take a keen interest in green assets and incorporate them into their portfolios to achieve optimal returns. Therefore, this article explores the static and dynamic connectedness between renewable energy stocks (solar, wind, and geothermal), green cryptocurrencies (Stellar, Nano, Cardona, and IOTA), and agricultural commodities (wheat, cocoa, coffee, corn, cotton, sugar, and soybean) using the TVP-VAR (time-varying parameter vector autoregression) framework offering novel empirical evidence for investors and portfolio managers. The connectedness is examined across two distinct sub-samples: during COVID-19 and post-COVID-19 times. Because the relevant connectedness can have implications for diversification benefits, we proceed with the computation of optimal weights, hedge ratios, and hedge effectiveness using the DCC-GARCH model. The main findings are as follows: We first find that green cryptocurrencies particularly Cardona and Stellar exhibit the highest spillovers to the network and wind energy stock has the least connectedness with the other markets. Second, the dynamic NET spillover indices reveal that cotton, cocoa, and coffee are consistently net receivers over the entire period except in the beginning of the pandemic. Third, renewable energy stocks exhibit diverse positions implying that the impact of the pandemic has varied significantly across the sectors. Finally, agricultural commodity depicts greater weights in the pandemic period under scoring the benefit of a diversified portfolio consisting of agriculture and green assets.

Aronia melanocarpa (black chokeberry) is a medicinally valuable small fruit species, yet its commercial propagation remains limited by low rooting and genotype-specific responses. This study developed an efficient, callus-free micropropagation and rooting protocol using a Shrub Plant Medium (SPM) supplemented with 5 mg/L BAP in large 660 mL jars, which yielded up to 27 shoots per explant. Optimal rooting (100%) was achieved with 0.5 mg/L NAA + 0.25 mg/L IBA in half-strength SPM. In the second phase, supervised machine learning models, including Random Forest (RF), XGBoost, Gaussian Process (GP), and Multilayer Perceptron (MLP), were employed to predict morphogenic traits based on culture conditions. XGBoost and RF outperformed other models, achieving R2 values exceeding 0.95 for key variables such as shoot number and root length. These results demonstrate that data-driven modeling can enhance protocol precision and reduce experimental workload in plant tissue culture. The study also highlights the potential for combining physiological understanding with artificial intelligence to streamline future in vitro applications in woody species.