Search Results (10)

Low-resolution (LR) imagery poses significant challenges to object detection systems, particularly in autonomous and resource-constrained environments where bandwidth and sensor quality are limited. To address this issue, this paper presents an integrated framework that enhances object detection performance by incorporating a Super-Resolution (SR) preprocessing stage prior to detection. Specifically, a Dense Residual Connected Transformer (DRCT) is employed to reconstruct high-resolution (HR) images from LR inputs, effectively restoring fine-grained structural and textural information essential for accurate detection. The reconstructed HR images are subsequently processed by a YOLOv11 detector without requiring architectural modifications. Experimental evaluations demonstrate consistent improvements across multiple scaling factors, with an average increase of 13.4% in Mean Average Precision (mAP)@50 at ×2 upscaling and 9.7% at ×4 compared with direct LR detection. These results validate the effectiveness of the proposed SR-based preprocessing approach in mitigating the adverse effects of image degradation. The proposed method provides an improved yet computationally challenging solution for object detection. Full article

Most existing deep learning-based super-resolution (SR) methods for remote sensing images rely on predefined degradation assumptions (e.g., bicubic downsampling). However, when real-world degradations deviate from these assumptions, their performance deteriorates significantly. Moreover, explicit degradation estimation approaches based on iterative schemes inevitably lead to accumulated estimation errors and time-consuming processes. In this paper, instead of explicitly estimating degradation types, we first innovatively introduce an MSCN_G coefficient to capture global prior information corresponding to different distortions. Subsequently, distortion-enhanced representations are implicitly estimated through contrastive learning and embedded into a super-resolution network equipped with multiple distortion decoders (D-Decoder). Furthermore, we propose a distortion-related channel segmentation (DCS) strategy that reduces the network’s parameters and computation (FLOPs). We refer to this Global Prior-guided Distortion-enhanced Representation Learning Network as GDRNet. Experiments on both synthetic and real-world remote sensing images demonstrate that our GDRNet outperforms state-of-the-art blind SR methods for remote sensing images in terms of overall performance. Under the experimental condition of anisotropic Gaussian blurring without added noise, with a kernel width of 1.2 and an upscaling factor of 4, the super-resolution reconstruction of remote sensing images on the NWPU-RESISC45 dataset achieves a PSNR of 28.98 dB and SSIM of 0.7656. Full article

Background and Objectives: Implant brand identification is critical in modern dental clinical diagnostics. With the increasing variety of implant brands and the difficulty of accurate identification in periapical radiographs, there is a growing demand for automated solutions. This study aims to leverage deep learning techniques to assist in dental implant classification, providing dentists with an efficient and reliable tool for implant brand detection. Methods: We proposed an innovative implant brand feature extraction method with multiple image enhancement techniques to improve implant visibility and classification accuracy. Additionally, we introduced a PA resolution enhancement technique that utilizes Dark Channel Prior and Lanczos interpolation for image resolution upscaling. Results: We evaluated the performance differences among various YOLO models for implant brand detection. Additionally, we analyzed the impact of implant brand feature extraction and PA resolution enhancement techniques on YOLO’s detection accuracy. Our results show that IB-YOLOv10 achieves a 17.8% accuracy improvement when incorporating these enhancement techniques compared to IB-YOLOv10 without enhancements. In real-world clinical applications, IB-YOLOv10 can classify implant brands in just 6.47 ms per PA, significantly reducing diagnostic time. Compared to existing studies, our model improves implant detection accuracy by 2.3%, achieving an overall classification accuracy of 94.5%. Conclusions: The findings of this study demonstrate that IB-YOLOv10 effectively reduces the diagnostic burden on dentists while providing a fast and reliable implant brand detection solution, improves clinical efficiency, and establishes a robust deep learning approach for automated implant detection in PA. Full article

The use of tungsten fiber-reinforced tungsten composites (W_f/W) has been demonstrated to significantly enhance the mechanical properties of tungsten (W) by incorporating W-fibers into the W-matrix. However, prior research has been restricted by the usage of single fiber-based textile fabrics, consisting of 150 µm warp and 50 µm weft filaments, with limited homogeneity, reproducibility, and mechanical properties in bulk structures due to the rigidity of the 150 µm W-fibers. To overcome this limitation, two novel textile preforms were developed utilizing radial braided W-yarns with 7 core and 16 sleeve filaments (R.B. 16 + 7), with a diameter of 25 µm each, as the warp material. In this study, bulk composites of two different fabric types were produced via a layer-by-layer CVD process, utilizing single 50 µm filaments (type 1) and R.B. 16 + 7 yarns (type 2) as weft materials. The produced composites were sectioned into KLST-type specimens based on DIN EN ISO 179-1:2000 using electrical discharge machining (EDM) and subjected to three-point bending tests. Both composites demonstrated enhanced mechanical properties with pseudo-ductile behavior at room temperature and withstood over 10,000 load cycles between 50–90% of their respective maximum load without sample fracture in three-point cyclic loading tests. Furthermore, a novel approach to predict the fatigue behavior of the material under cyclic loading was developed based on the high reproducibility of the composites produced, especially for the composite based on type 1. This approach provides a new benchmark for upscaling endeavors and may enable a better prediction of the service life of the produced components made of W_f/W in the future. In comparison, the composite based on fabric type 1 demonstrated superior results in manufacturing performance and mechanical properties. With a high relative average density (>97%), a high fiber volume fraction (14–17%), and a very homogeneous fiber distribution in the CVD-W matrix, type 1 shows a promising option to be further tested in high heat flux tests and to be potentially used as an alternative to currently used materials for the most stressed components of nuclear fusion reactors or other potential application fields such as concentrated solar power (CSP), aircraft turbines, the steel industry, quantum computing, or welding tools. Type 2 composites have a higher layer spacing compared to type 1, resulting in gaps within the matrix and less homogeneous material properties. While type 2 composites have demonstrated a notable enhancement over 150 µm fiber-based composites, they are not viable for industrial scale-up unlike type 1 composites. Full article

As excellent biocontrol agents and plant growth promoters, Trichoderma species are agriculturally important. Trichoderma spp. cultures can be produced using solid-state or submerged cultivation, the latter being much less labor intensive and easier to control and automate. The aim of the study was to investigate the ability to increase the shelf-life of T. asperellum cells by optimizing cultivation media and upscaling the submerged cultivation process. Four different cultivation media were used with or without the addition of Tween 80 and stored with or without incorporation into peat, and viability, expressed as CFU/g, was assessed during one year of storage in an industrial warehouse. The addition of Tween 80 had a positive effect on the biomass yield. The culture medium played a major role in the ability of the mycelium to produce spores, which in turn influenced the amount of CFU. This effect was less pronounced when the biomass was mixed with peat prior to storage. A procedure that increases the number of CFU in a peat-based product formulation is recommended, namely, incubation of the mixture at 30 °C for 10 days prior to storage at 15 °C over an extended period of time. Full article

MR scans of low-gamma X-nuclei, low-concentration metabolites, or standard imaging at very low field entail a challenging tradeoff between resolution, signal-to-noise, and acquisition duration. Deep learning (DL) techniques, such as UNets, can potentially be used to improve such “low-quality” (LQ) images. We investigate three UNets for upscaling LQ MRI: dense (DUNet), robust (RUNet), and anisotropic (AUNet). These were evaluated for two acquisition scenarios. In the same-subject High-Quality Complementary Priors (HQCP) scenario, an LQ and a high quality (HQ) image are collected and both LQ and HQ were inputs to the UNets. In the No Complementary Priors (NoCP) scenario, only the LQ images are collected and used as the sole input to the UNets. To address the lack of same-subject LQ and HQ images, we added data from the OASIS-1 database. The UNets were tested in upscaling 1/8, 1/4, and 1/2 undersampled images for both scenarios. As manifested by non-statically significant differences of matrices, also supported by subjective observation, the three UNets upscaled images equally well. This was in contrast to mixed effects statistics that clearly illustrated significant differences. Observations suggest that the detailed architecture of these UNets may not play a critical role. As expected, HQCP substantially improves upscaling with any of the UNets. The outcomes support the notion that DL methods may have merit as an integral part of integrated holistic approaches in advancing special MRI acquisitions; however, primary attention should be paid to the foundational step of such approaches, i.e., the actual data collected. Full article

Fish is a good source of nutrients, although it is easily spoiled. As such, drying is a common method of preserving fish to compensate for its perishability. Dried fish exists in different cultures with varying types of fish used and drying methods. These delicacies are not only consumed for their convenience and for their health benefits, as discussed in this review. Most commonly, salt and spices are added to dried fish to enhance the flavours and to decrease the water activity (a_w) of the fish, which further aids the drying process. For fish to be dried effectively, the temperature, drying environment, and time need to be considered along with the butchering method used on the raw fish prior to drying. Considering the various contributing factors, several physicochemical and biochemical changes will certainly occur in the fish. In this review, the pH, water activity (a_w), lipid oxidation, and colour changes in fish drying are discussed as well as the proximate composition of dried fish. With these characteristic changes in dried fish, the sensory, microbial and safety aspects of dried fish are also affected, revolving around the preferences of consumers and their health concerns, especially based on how drying is efficient in eliminating/reducing harmful microbes from the fish. Interestingly, several studies have focused on upscaling the efficiency of dried fish production to generate a safer line of dried fish products with less effort and time. An exploratory approach of the published literature was conducted to achieve the purpose of this review. This evaluation gathers important information from all available library databases from 1990 to 2022. In general, this review will benefit the fishery and food industry by enabling them to enhance the efficiency and safety of fish drying, hence minimising food waste without compromising the quality and nutritional values of dried fish. Full article

Upscaling an operation typically results in economies of scale, i.e., cost advantages in business, especially when the production unit’s utilization rate can be improved. According to economic studies of mechanized timber harvesting, large wood harvesting entrepreneurs tend to be more successful in business than small entrepreneurs. What are the factors that influence harvesting costs, and how great is their effect on costs? These questions were investigated in mechanized cut-to-length timber harvesting in Eastern Finland by varying (a) the size of the harvesting fleet, (b) the harvesting site reserve, and (c) the timing and duration of the working day of machine relocations, in the case of an entrepreneur using a discrete-event simulation method. Prior to the simulations, harvesting site data were generated from the National Forest Inventory data by the MELA software, and the spatial data analyses by ArcGIS. According to the results, largely because of the low utilization rate of the contractor’s own relocation truck, the harvesting cost of a 2-harvesting-unit (2 HU) scenario was 9% or 6% higher than 4 HU, and 13% or 8% higher than 8 HU, with or without a specifically employed driver of a relocation truck, respectively (the harvesting unit consists of a harvester and a forwarder). In the 4 and 8 HU scenarios, harvesting costs decreased on average by 1% (0.3–1.5), when doubling the size of the harvesting site reserve. With fleet sizes of 6 and 8 HU, good utilization of a relocation truck reduced relocation costs, whereas machine costs only increased a small amount because of a longer machine relocation waiting time than with smaller entrepreneurs. The study raised the importance of entrepreneur-specific planning of machine relocations in the cost-efficient timber harvesting in Finland. Full article

Hyperspectral image (HSI) super-resolution (SR) is a challenging task due to its ill-posed nature, and has attracted extensive attention by the research community. Previous methods concentrated on leveraging various hand-crafted image priors of a latent high-resolution hyperspectral (HR-HS) image to regularize the degradation model of the observed low-resolution hyperspectral (LR-HS) and HR-RGB images. Different optimization strategies for searching a plausible solution, which usually leads to a limited reconstruction performance, were also exploited. Recently, deep-learning-based methods evolved for automatically learning the abundant image priors in a latent HR-HS image. These methods have made great progress for HS image super resolution. Current deep-learning methods have faced difficulties in designing more complicated and deeper neural network architectures for boosting the performance. They also require large-scale training triplets, such as the LR-HS, HR-RGB, and their corresponding HR-HS images for neural network training. These training triplets significantly limit their applicability to real scenarios. In this work, a deep unsupervised fusion-learning framework for generating a latent HR-HS image using only the observed LR-HS and HR-RGB images without previous preparation of any other training triplets is proposed. Based on the fact that a convolutional neural network architecture is capable of capturing a large number of low-level statistics (priors) of images, the automatic learning of underlying priors of spatial structures and spectral attributes in a latent HR-HS image using only its corresponding degraded observations is promoted. Specifically, the parameter space of a generative neural network used for learning the required HR-HS image to minimize the reconstruction errors of the observations using mathematical relations between data is investigated. Moreover, special convolutional layers for approximating the degradation operations between observations and the latent HR-HS image are specifically to construct an end-to-end unsupervised learning framework for HS image super-resolution. Experiments on two benchmark HS datasets, including the CAVE and Harvard, demonstrate that the proposed method can is capable of producing very promising results, even under a large upscaling factor. Furthermore, it can outperform other unsupervised state-of-the-art methods by a large margin, and manifests its superiority and efficiency. Full article

The synthesis of Mg-Al, Ca-Al, Zn-Al and Cu-Al layered double hydroxides (LDHs) was investigated with a one-step wet mechanochemical route. The research aims to expand on the mechanochemical synthesis of LDH using a mill designed for wet grinding application. A 10% slurry of solids was added to a Netzsch LME 1 horizontal bead mill and milled for 1 h at 2000 rpm. Milling conditions were selected according to machine limitations and as an initial exploratory starting point. Precursor materials selected consisted of a mixture of oxides, hydroxides and basic carbonates. Samples obtained were divided such that half was filtered and dried at 60 °C for 12 h. The remaining half of the samples were further subjected to ageing at 80 °C for 24 h as a possible second step to the synthesis procedure. Synthesis conditions, such as selected precursor materials and the M^II:M^III ratio, were adapted from existing mechanochemical methods. LDH synthesis prior to ageing was successful with precursor materials observably present within each sample. No Cu-Al LDH was clearly identifiable. Ageing of samples resulted in an increase in the conversion of raw materials to LDH product. The research offers a promising ‘green’ method for LDH synthesis without the production of environmentally harmful salt effluent. The synthesis technique warrants further exploration with potential for future commercial up-scaling. Full article