5916 research outputs found
Sort by
Shoplifting detection from customer behavior using deep learning
No full textThe increase in shoplifting in the retail market causes significant stock and profit losses. Existing security methods are often costly, prone to human error, and not applicable to all product types, highlighting the need for an innovative, low-cost, and effective solution against theft. This study presents a deep learning-based system for detecting shoplifting behavior from surveillance video footage. The system integrates four components: (1) person detection to identify customers approaching shelves, (2) activity recognition to analyze movements for suspicious behavior, (3) product detection to determine which items are taken, and (4) person re-identification model which matches suspicious customers when they arrive at the checkout were developed. A Time Distributed CNN-LSTM model was developed for activity recognition; YOLOv4 was fine-tuned for person and product detection, and Siamese Networks were used for person re-identification. Training and testing were conducted using a data set collected from both an office demo setup and a real retail environment, covering five different shoplifting scenarios. The dataset collected includes 1219 videos across five scenarios. The proposed system was evaluated on a custom-collected dataset and achieved 95% overall accuracy, with component-level accuracy of 85% for activity recognition, 97% for person and product detection, and 87% for person re-identification. In this paper, the authors suggested a model which primarily focuses on recognizing shoplifting actions. The originality of this study lies in the integrated system, including 4 components; person detection, activity recognition, product detection and person re-identification which work simultaneously to provide end-to-end solutions.TÜBİTA
Maternal symptoms and emotional availability predicting children's behavior problems: A longitudinal study
No full textThe longitudinal research focusing on the effects of maternal mental health on parenting capacity and child behavior problems during COVID-19 is still limited. Therefore, we examined how maternal symptoms of anxiety, depression, stress, and COVID-19-related stress affect maternal emotional availability at a 2-month follow-up and behavior problems at a 4-month follow-up. The mothers (N = 443) with pre-schoolers ( Age Range = 30-80 months) responded to questionnaires at three time points. Maternal depression and stress positively predicted maternal hostility and negatively predicted maternal mutual attunement. Maternal hostility and mutual attunement mediated the associations between maternal depression and child externalization, as well as maternal stress and child externalization. Only hostility mediated the associations between maternal depression and stress with child internalization. We highlighted the role of maternal depression and stress in both positive and negative aspects of emotional availability that could possibly shape child externalization. For internalization, maternal hostility adopted the mediator role.TÜBİTA
Hstr-net: Reference based video super-resolution with dual cameras
No full textHigh-spatio-temporal resolution (HSTR) video recording plays a crucial role in enhancing various imagery tasks that require fine-detailed information. State-of-the-art cameras provide this required high frame-rate and high spatial resolution together, albeit at a high cost. To alleviate this issue, this paper proposes a dual camera system for the generation of HSTR video using reference-based super-resolution (RefSR). One camera captures high spatial resolution low frame rate (HSLF) video while the other captures low spatial resolution high frame rate (LSHF) video simultaneously for the same scene. A novel deep learning architecture is proposed to fuse HSLF and LSHF video feeds and synthesize HSTR video frames. The proposed model combines optical flow estimation and (channel-wise and spatial) attention mechanisms to capture the fine motion and complex dependencies between frames of the two video feeds. Simulations show that the proposed model provides significant improvement over existing reference-based SR techniques in terms of PSNR and SSIM metrics. The method also exhibits sufficient frames per second (FPS) for aerial monitoring when deployed on a power-constrained drone equipped with dual cameras. The source code is publicly available at https://github.com/umutsuluhan/HSTRNet. © The Author(s) 2025.TÜBİTAKPublisher versio
Driving digital transformation: marketing solutions for servitization challenges a paper written with ChatGPT
No full textThis report examines the critical role of marketing strategies in enabling the successful transition of manufacturing firms from product-centric models to integrated product-service systems—a transformation known as servitization. Drawing on qualitative analysis of case studies and published
interviews, this study identifies key gaps in marketing approaches, including the challenges of communicating value, balancing customization with scalability, leveraging digital tools, and adapting strategies to diverse market needs. By addressing these gaps, the report proposes actionable
solutions informed by theoretical frameworks such as Service-Dominant Logic (SDL), Relationship Marketing, and Integrated Marketing Communications (IMC).
Findings highlight the importance of clear value propositions, modular service systems, and personalized marketing campaigns supported by digital technologies like IoT and predictive analytics. Case studies from industry leaders such as Rolls-Royce, Siemens, and Caterpillar illustrate successful applications, while examples from Kodak and IBM underscore the consequences of inadequate marketing in servitization efforts. The discussion further emphasizes the
interconnectedness of identified gaps and proposes an integrated approach to align marketing strategies with operational capabilities and customer expectations.
This research contributes to academic discourse by bridging theoretical insights with practical applications, offering a nuanced understanding of servitization marketing. It provides actionable recommendations for manufacturers to enhance customer engagement, foster adoption, and sustain competitive advantage. Limitations and future research directions are discussed, highlighting opportunities for quantitative validation and cross-industry exploration. This study underscores marketing’s pivotal role in driving the digital transformation of servitized business models, positioning it as a cornerstone for sustainable growth in a rapidly evolving global landscape
Mikroakışkan temelli floresans mikroskop sistemi ile otofloresan flavin koenzimlerinin fotofiziksel geçişleri üzerine nümerik modelleme çalışmaları
No full textBu araştırmada ışığa ve çözücü ortamına oldukça hassas, çok küçük uyarım kesit alanına sahip ve zayıf floresan ışıma yapabilen flavin mononükleotit (FMN) ve flavin adenin dinükleotit (FAD) koenzimlerinin fotofiziksel geçişlerini çözümleme kapasitesine sahip mikroakışkan temelli bir floresans mikroskop sistemi için nümerik modelleme çalışmaları sunulmuştur. FMN ve FAD’nin moleküler yapısı, fotofiziksel özellikleri ve girdikleri kimyasal reaksiyonlar dikkate alınarak her iki molekül için farklı fotofiziksel modeller kullanılmıştır. Bu modellerde yer alan elektronik durumlar 1. mertebeden lineer diferansiyel denklem sistemi olarak ele alınmış olup her bir elektronik durum popülasyonu zamana bağlı olarak çözülmüş, mikroakışkan çip ile lazer uyarım alanının geometrik boyutları ve mikroskop parametreleri kullanılarak görüntü ve sinyal verisi olarak elde edilmiştir. İki farklı akış hızında lazer uyarım şiddeti, çözücüye eklenen etanol, askorbat ve triptofan gibi redoks ajanlarının normalize floresan sinyaline ve elektronik durum popülasyonlarına olan etkisi simüle edilmiştir. Sinyal ve elektronik durum analizlerine ek olarak sinyallerin oluşturulmasında kullanılan sCMOS görüntü verileri farklı deneysel koşullar için simüle edilmiş ve lazer uyarım alanıyla kıyaslanmıştır. Araştırmada önerilen yöntem farklı akış hızlarında farklı karanlık durum popülasyonlarının birbirinden ayırt edilebilirliğini ve farklı deneysel koşullarda değişen karanlık durumların normalize floresan sinyaline ve kamera görüntülerine olan etkisini çözümleme kapasitesine sahip olduğunu göstermiştir. Mevcut yöntemlerle kıyaslandığında, elde edilen sayısal bulgular, çalışmada sunulan yöntemin flavin foto-bozunumunu büyük ölçüde önleyebilme potansiyelini ispatlamıştır ve farklı moleküllerin fotofiziksel özelliklerinin hangi koşullarda gözlemlenebileceği ile ilgili optimizasyon çalışmalarının yapılmasına olanak sağlamaktadır.TÜBİTAKPublisher versio
On the hot workability of Ti-6Al-4V based on thermal processing maps and artificial neural network modeling
No full textFor probing the hot workability of Ti-6Al-4V alloy, uniaxial tensile tests were carried out in the temperature range of 500-800 degrees C and at quasi-static strain rates. The effects of deformation temperature and strain rate on the flow behavior of Ti-6Al-4V were investigated. The flow behavior was modeled with the artificial neural network approach with the resulting correlation coefficient and average absolute relative error of 0.9997 and 2.90621%, respectively. The statistical evaluation results showed that the applied model could predict with very high reliability for the range of deformation parameters. The 3D thermal processing maps indicate that there are unsafe conditions at low temperature and high strain rates. In addition, in cases where the temperature is high and the strain rate is low, flow stability is observed and high-power distribution efficiency is determined. The validation of thermal processing maps was carried out by microstructural examinations and fractographic analysis. Overall, it is asserted that the temperature range of 700-800 degrees C and the strain rates of 0.01 and 0.001 s-1 are the optimized process parameters for hot forming of Ti-6Al-4V.Özyeğin Üniversitesi ; TÜBİTAK ; Scientific Research Program of Turkish Aerospace Industrie
Atomistic investigation of porous amorphous materials for CH4/H2 separation
No full textRevealing the gas separation capabilities of amorphous porous materials remains a critical challenge in the materials community for their development as novel adsorbents. This work aims to unlock the potential of amorphous materials for adsorption-based CH4/H-2 separation at pressure swing adsorption (PSA) condition using grand canonical Monte Carlo (GCMC) simulations. Several adsorbent performance evaluation metrics, including adsorption selectivity, working capacity, adsorbent performance score (APS) and regenerability (R%) were computed at 298 K for polymers of intrinsic microporosity (PIMs), amorphous carbons, kerogens, and amorphous zeolitic imidazole frameworks (ZIFs). The CH4/H-2 selectivities and CH4 working capacities of the amorphous materials were estimated to be 9-62 and 0.1-5 mol/kg under PSA condition. Kerogens exhibited the highest APS, and most of the structures provided high R%>80 %. However, none of the materials could reach the maximum APS (802 mol/kg) of crystalline MOFs. Diffraction pattern analysis of crystalline and amorphous ZIF-4 was also performed, and the structural changes were monitored to independently confirm the amorphization. Although crystalline ZIFs exhibited higher adsorption selectivities for CH4/H-2 separation than amorphous ZIFs, their R% were significantly lower. Gas mixture adsorption isotherms of promising amorphous materials were also computed to reveal gas adsorption mechanism. The developed computational approach will be useful in predicting the performance of amorphous materials for CH4/H-2 separation under industrial conditions and monitoring amorphization by diffraction analysis during mass production.TÜBİTA
BDDK kararlarının kur beklentilerine etkisi.
No full textRisk-neutral distributions (RND) of currency options are useful for predicting the price movements of future exchange rates, valuing financial derivatives, and applying appropriate monetary policy. This study examines the swap restriction regulations applied by the BRSA to prevent USD/TRY exchange rate shocks after the diplomatic crisis with the USA. These regulations restricted offshore swap transactions in which Turkish banks give foreign currency and receive Turkish Lira to foreign banks at maturity. Risk-neutral distributions (RND) are obtained for all maturities using the non-parametric Malz approach. The RND findings obtained indicate a deterioration in the expectations regarding the USD/TRY exchange rate for all maturities following the swap restrictions. In addition to the visual representation, sharp movements in exchange rate expectations are displayed with moments of RNDs up to the 4th degree. Moreover, the impact capacity of regulatory decisions taken by the BRSA, the moments of RNDs up to the 4th degree, and the relationship between global factors is examined with the established model. The findings indicate that the effect of the regulations is temporary.Döviz opsiyonlarının riskten bağımsız dağılımları (RND), gelecekteki döviz kurlarının fiyat hareketlerini tahmin etmek, finansal türevlerin değerlemesini yapmak ve uygun para politikasını uygulamak için kullanışlıdır. Bu çalışma, ABD ile yaşanan diplomatik kriz sonrasında Bankacılık Düzenleme ve Denetleme Kurulu'nun (BDDK) USD/TRY kur şoklarına önlemek için uyguladığı swap kısıtlama düzenlemelerini incelemektedir. Bu düzenlemeler, Türk bankalarının vade sonunda yabancı bankalara döviz verip Türk Lirası aldığı offshore swap işlemlerine sınırlama getirmiştir. Parametrik olmayan Malz yaklaşımı kullanılarak tüm vadeler için riske duyarsız dağımlımlar (RND) elde edilmiştir. Elde edilen RND bulguları, alınan kararların ardından tüm vadeler için USD/TL kuruna ilişkin beklentilerde bozulmaya işaret etmektedir. Görsel temsilin yanı sıra 4. Dereceye kadar olan RND momentleri ile kur beklentilerine ilişkin keskin hareketler göz önüne serilmiştir. Ayrıca, BDDK tarafından alınan düzenleyici kararların etki kapasitesi, 4. dereceye kadar RND momentleri ve küresel faktörler arasındaki ilişki kurulan model ile incelenmektedir. Bulgular düzenlemelerin etkisinin geçici olduğuna işaret etmektedir
Entrepreneurial back-to-landers: Neo-farmers in Turkey
No full textUrban-to-rural migration, particularly back-to-land migration, has become prominent in Turkey. This paper focuses on entrepreneurial back-to-landers or neo-farmers, who have migrated to rural areas specifically to get into commercial agriculture and farming. Through an analysis of semi-structured interviews with 72 neofarmers, the paper identifies six critical attributes that aid them during entry into farming and later on, when they are running successful farm businesses: possession of financial wealth; ownership of agricultural land; familiarity with agro-food sectors; education (in agricultural and/or food sciences); experience with corporate conduct; and active connections to the local and/or national organizations associated with the food movement. The paper argues that these attributes feature entrepreneurial skills, experience and connections, which in turn provide neo-farmers with economic, social, and cultural capital and comparative advantages to run their farm businesses. Through the case, the paper shows that one, the direction of capital flows, which historically has been rural-to-urban, may change to urban-to-rural (and from non-agricultural sectors to agriculture) through entrepreneurial back-to-land migration; and two, entrepreneurial skills have become vital for smallholders - newcomer or continuer - if agriculture is going to be their primary income generating activity.TÜBİTA
Bee bread incorporated chitosan films for food packaging: Enhanced physical properties, antioxidant, and antibacterial activities
No full textBiodegradable food packaging films have garnered significant interest in recent years due to environmental concerns and the need for enhanced food security. This study reports, for the first time, the preparation and characterization of chitosan film incorporated with bee bread extract, focusing on its physical, antioxidant, and antibacterial properties. Bee bread, a fermented bee product rich in bioactive compounds, offers nutritional and therapeutic benefits, making it a promising natural additive for functional food packaging materials. The films were prepared by the solvent-casting method. Fourier transform infrared spectroscopy (FTIR) confirmed the interactions between bee bread extract and chitosan. Scanning electron microscopy (SEM) analyses demonstrated strong compatibility of bee bread extract within the chitosan. The incorporation of bee bread extract significantly (p < 0.05) improved mechanical properties, such as tensile strength, while also markedly enhancing antibacterial activity. Gluconic acid, which is the dominant organic acid in bee bread, was released at a rate of 32% from the films within the first three hours. The 2,2-diphenyl-1-picrylhydrazyl (DPPH) radical scavenging assay showed a substantial increase in the bee bread-incorporated film (88.36%) compared to the pure chitosan film (5.42%). The film's transparency to UV and visible light, as well as its color analysis and physical characteristics, were thoroughly evaluated. Furthermore, the use of the film in food preservation demonstrated notable effectiveness, with the samples maintaining a fresh appearance throughout the storage period. Overall, bee bread extract presents a promising natural additive for enhancing chitosan-based films by improving mechanical strength, antioxidant capacity, and antibacterial effectiveness.TÜBİTAK ; Istanbul University ; Istanbul Technical Universit