Search Results (44)

Voice conversion (VC) refers to the technique of modifying one speaker’s voice to mimic another’s while retaining the original linguistic content. This technology finds its applications in fields such as speech synthesis, accent modification, medicine, security, privacy, and entertainment. Among the various deep generative models used for voice conversion, including variational autoencoders (VAEs) and generative adversarial networks (GANs), diffusion models (DMs) have recently gained attention as promising methods due to their training stability and strong performance in data generation. Nevertheless, traditional DMs focus mainly on learning reconstruction paths like VAEs, rather than conversion paths as GANs do, thereby restricting the quality of the converted speech. To overcome this limitation and enhance voice conversion performance, we propose a cycle-consistent diffusion (CycleDiffusion) model, which comprises two DMs: one for converting the source speaker’s voice to the target speaker’s voice and the other for converting it back to the source speaker’s voice. By employing two DMs and enforcing a cycle consistency loss, the CycleDiffusion model effectively learns both reconstruction and conversion paths, producing high-quality converted speech. The effectiveness of the proposed model in voice conversion is validated through experiments using the VCTK (Voice Cloning Toolkit) dataset. Full article

In recent years, advancements in high-resolution digital twin platforms or artificial intelligence marine forecasting have led to the increased requirements of high-resolution oceanic data. However, existing sea surface temperature (SST) products from observations often fail to meet researchers’ resolution requirements. Deep learning models serve as practical techniques for improving the spatial resolution of SST data. In particular, diffusion models (DMs) have attracted widespread attention due to their ability to generate more vivid and realistic results than other neural networks. Despite DMs’ potential, their application in SST spatial downscaling remains largely unexplored. Hence we propose a novel DM-based spatial downscaling model, called DIFFDS, designed to obtain a high-resolution version of the input SST and to restore most of the meso scale processes. Experimental results indicate that DIFFDS is more effective and accurate than baseline neural networks, its downscaled high-resolution SST data are also visually comparable to the ground truth. The DIFFDS achieves an average root-mean-square error of 0.1074 °C and a peak signal-to-noise ratio of 50.48 dB in the 4× scale downscaling task, which shows its accuracy. Full article

Accurate skin diagnosis through end-user applications is important for early detection and cure of severe skin diseases. However, the low quality of dermoscopic images hampers this mission, especially with the presence of hair on these kinds of images. This paper introduces DM–AHR, a novel, self-supervised conditional diffusion model designed specifically for the automatic generation of hairless dermoscopic images to improve the quality of skin diagnosis applications. The current research contributes in three significant ways to the field of dermatologic imaging. First, we develop a customized diffusion model that adeptly differentiates between hair and skin features. Second, we pioneer a novel self-supervised learning strategy that is specifically tailored to optimize performance for hairless imaging. Third, we introduce a new dataset, named DERMAHAIR (DERMatologic Automatic HAIR Removal Dataset), that is designed to advance and benchmark research in this specialized domain. These contributions significantly enhance the clarity of dermoscopic images, improving the accuracy of skin diagnosis procedures. We elaborate on the architecture of DM–AHR and demonstrate its effective performance in removing hair while preserving critical details of skin lesions. Our results show an enhancement in the accuracy of skin lesion analysis when compared to existing techniques. Given its robust performance, DM–AHR holds considerable promise for broader application in medical image enhancement. Full article

Surface defect inspection based on deep learning has demonstrated outstanding performance in improving detection accuracy and model generalization. However, the small scale of defect datasets always limits the application of deep models in industry. Generative models can obtain realistic samples in a very cheap way, which can effectively solve this problem and thus has received widespread attention in recent years. This paper provides a comprehensive analysis and summary of the current studies of surface defect inspection methods proposed between 2022 and 2024. First, according to the use of generative models, these methods are classified into four categories: Variational Auto-Encoders (VAEs), Generative Adversarial Networks (GANs), Diffusion Models (DMs), and multi-models. Second, the research status of surface defect inspection based on generative models in recent years is discussed from four aspects: sample generation, detection objective, inspection task, and learning model. Then, the public datasets and evaluation metrics that are commonly used for surface defect inspection are discussed, and a comparative evaluation of defect inspection methods based on generative models is provided. Finally, this study discusses the existing challenges for the defect inspection methods based on generative models, providing insights for future research. Full article

Universal image restoration (UIR) aims to accurately restore images with a variety of unknown degradation types and levels. Existing methods, including both learning-based and prior-based approaches, heavily rely on low-quality image features. However, it is challenging to extract degradation information from diverse low-quality images, which limits model performance. Furthermore, UIR necessitates the recovery of images with diverse and complex types of degradation. Inaccurate estimations further decrease restoration performance, resulting in suboptimal recovery outcomes. To enhance UIR performance, a viable approach is to introduce additional priors. The current UIR methods have problems such as poor enhancement effect and low universality. To address this issue, we propose an effective framework based on a diffusion model (DM) for universal image restoration, dubbed ETDiffIR. Inspired by the remarkable performance of text prompts in the field of image generation, we employ text prompts to improve the restoration of degraded images. This framework utilizes a text prompt corresponding to the low-quality image to assist the diffusion model in restoring the image. Specifically, a novel text–image fusion block is proposed by combining the CLIP text encoder and the DA-CLIP image controller, which integrates text prompt encoding and degradation type encoding into time step encoding. Moreover, to reduce the computational cost of the denoising UNet in the diffusion model, we develop an efficient restoration U-shaped network (ERUNet) to achieve favorable noise prediction performance via depthwise convolution and pointwise convolution. We evaluate the proposed method on image dehazing, deraining, and denoising tasks. The experimental results indicate the superiority of our proposed algorithm. Full article

Generative adversarial networks (GANs) and diffusion models (DMs) have revolutionized the creation of synthetically generated but realistic-looking images. Distinguishing such generated images from real camera captures is one of the key tasks in current multimedia forensics research. One particular challenge is the generalization to unseen generators or post-processing. This can be viewed as an issue of handling out-of-distribution inputs. Forensic detectors can be hardened by the extensive augmentation of the training data or specifically tailored networks. Nevertheless, such precautions only manage but do not remove the risk of prediction failures on inputs that look reasonable to an analyst but in fact are out of the training distribution of the network. With this work, we aim to close this gap with a Bayesian Neural Network (BNN) that provides an additional uncertainty measure to warn an analyst of difficult decisions. More specifically, the BNN learns the task at hand and also detects potential confusion between post-processing and image generator artifacts. Our experiments show that the BNN achieves on-par performance with the state-of-the-art detectors while producing more reliable predictions on out-of-distribution examples. Full article

In this short review, corresponding to a talk given at the conference “Cosmology 2023 in Miramare”, we combine an analysis of five regions observed by H.E.S.S. in the Galactic Center, intending to constrain the Dark Matter (DM) density profile in a WIMP annihilation scenario. For the analysis, we include the state-of-the-art Galactic diffuse emission Gamma-optimized model computed with DRAGON and a wide range of DM density profiles from cored to cuspy profiles, including different kinds of DM spikes. Our results are able to constrain generalized NFW profiles with an inner slope $\gamma \gtrsim 1.3$. When considering DM spikes, the adiabatic spike is completely ruled out. However, smoother spikes given by the interactions with the bulge stars are compatible if $\gamma \lesssim 0.8$, with an internal slope of ${\gamma }_{\mathrm{sp-stars}}=1.5$. Full article

Synthetic aperture radar (SAR) plays a crucial role in maritime surveillance due to its capability for all-weather, all-day operation. However, SAR ship recognition faces challenges, primarily due to the imbalance and inadequacy of ship samples in publicly available datasets, along with the presence of numerous outliers. To address these issues, this paper proposes a SAR ship classification method based on text-generated images to tackle dataset imbalance. Firstly, an image generation module is introduced to augment SAR ship data. This method generates images from textual descriptions to overcome the problem of insufficient samples and the imbalance between ship categories. Secondly, given the limited information content in the black background of SAR ship images, the Tokens-to-Token Vision Transformer (T2T-ViT) is employed as the backbone network. This approach effectively combines local information on the basis of global modeling, facilitating the extraction of features from SAR images. Finally, a Squeeze-and-Excitation (SE) model is incorporated into the backbone network to enhance the network’s focus on essential features, thereby improving the model’s generalization ability. To assess the model’s effectiveness, extensive experiments were conducted on the OpenSARShip2.0 and FUSAR-Ship datasets. The performance evaluation results indicate that the proposed method achieves higher classification accuracy in the context of imbalanced datasets compared to eight existing methods. Full article

This study aimed to evaluate the effect of convective and microwave drying on the bioactive-compounds content of blackberry (Rubus fruticosus) fruits, as well as drying parameters and energy consumption. The fruit was dehydrated in a convective dehydrator at a temperature of 50 °C and 70 °C and in a microwave oven at power levels of 90 W, 180 W and 240 W. The highest amount of anthocyanins, polyphenols and antioxidant capacity were obtained in blackberry fruits that were microwave dried at 90 W and 180 W (46.3–52.5 and 51.8–83.5 mg 100 g⁻¹ dm of total anthocyanins, 296.3–255.8 and 418.4–502.2 mg 100 g⁻¹ dm of total phenolics, and 1.20–1.51 and 1.45–2.35 mmol TE 100 g⁻¹ dm of antioxidant capacity for 90 W and 180 W models, respectively). It turned out that microwave dehydration shortened the processing time and lowered the energy consumption compared to convective drying (a significantly reduced drying time of 92–99% with microwave dehydration). Blackberry fruits dehydrated at 240 W showed the shortest dehydration time (59–67 min), minimal energy consumption (0.23 kWh) and the most efficient diffusion (1.48–1.66 × 10⁻⁸ m² s⁻¹). Full article

This study examined the convective drying of red cabbage at temperatures ranging from 50 to 90 °C. Mathematical modeling was used to describe isotherms, drying kinetics and rehydration process. The effects of drying conditions on energy consumption and microstructure were also evaluated. The Halsey model had the best fit to the isotherm data and the equilibrium moisture was determined to be 0.0672, 0.0490, 0 0.0379, 0.0324 and 0.0279 g water/g d.m. at 50, 60, 70, 80 and 90 °C, respectively. Drying kinetics were described most accurately by the Midilli and Kuçuk model. Also, the diffusion coefficient values increased with drying temperature. Lower energy consumption was found for drying at 90 °C and the rehydration process was best described by the Weibull model. Samples dehydrated at 90 °C showed high water holding capacity and better maintenance of microstructure. These results could be used to foster a sustainable drying process for red cabbage. Full article

Deepfakes are notorious for their unethical and malicious applications to achieve economic, political, and social reputation goals. Recent years have seen widespread facial forgery, which does not require technical skills. Since the development of generative adversarial networks (GANs) and diffusion models (DMs), deepfake generation has been moving toward better quality. Therefore, it is necessary to find an effective method to detect fake media. This contemporary survey provides a comprehensive overview of several typical facial forgery detection methods proposed from 2019 to 2023. We also analyze and group them into four categories in terms of their feature extraction methods and network architectures: traditional convolutional neural network (CNN)-based detection, CNN backbone with semi-supervised detection, transformer-based detection, and biological signal detection. Furthermore, it summarizes several representative deepfake detection datasets with their advantages and disadvantages. Finally, we evaluate the performance of these detection models with respect to different datasets by comparing their evaluating metrics. Across all experimental results on these state-of-the-art detection models, we find that the accuracy is largely degraded if we utilize cross-dataset evaluation. These results will provide a reference for further research to develop more reliable detection algorithms. Full article

We address the problem of data augmentation in a rotating turbulence set-up, a paradigmatic challenge in geophysical applications. The goal is to reconstruct information in two-dimensional (2D) cuts of the three-dimensional flow fields, imagining spatial gaps present within each 2D observed slice. We evaluate the effectiveness of different data-driven tools, based on diffusion models (DMs), a state-of-the-art generative machine learning protocol, and generative adversarial networks (GANs), previously considered as the best-performing method both in terms of point-wise reconstruction and the statistical properties of the inferred velocity fields. We focus on two different DMs recently proposed in the specialized literature: (i) RePaint, based on a heuristic strategy to guide an unconditional DM for flow generation by using partial measurements data, and (ii) Palette, a conditional DM trained for the reconstruction task with paired measured and missing data. Systematic comparison shows that (i) DMs outperform the GAN in terms of the mean squared error and/or the statistical accuracy; (ii) Palette DM emerges as the most promising tool in terms of both point-wise and statistical metrics. An important property of DMs is their capacity for probabilistic reconstructions, providing a range of predictions based on the same measurements, enabling uncertainty quantification and risk assessment. Full article

Purified red phosphorus (RP) can be used as an adsorbent. However, the adsorption mechanism and reuse ability of purified RP have not been reported. This study utilized X-ray diffraction, Fourier transform infrared spectroscopy and scanning electron microscopy techniques (a statistical physics model and the standard molar free energy of formation) to investigate the adsorption mechanism of methylene blue (MB) by purified RP. Purification did not change commercial RP structure according to X-ray diffraction. The results showed that the adsorption process only included physical adsorption according to Fourier transform infrared spectroscopy and UV–vis diffuse reflection absorption spectra. The specific areas of commercial RP and purified RP were 0.02 cm³/g and 5.27 cm³/g, respectively. Thus, purified RP has a higher adsorption capacity compared with commercial RP. A statistical physics model showed that, as the temperature increased from 288 to 308 K, the q_e, D_m and q_sat of purified RP for MB increased from 179.87, 0.824 and 0.824 to 303.26 mg/g, 1.497 mol/kg and 1.497 mol/kg, respectively. The fitted values of Δ_rS_m^θ, Δ_rH_m^θ and Δ_rG_m^θ were 104.38 J·mol⁻¹·K⁻¹, −2.7 × 10³ J·mol⁻¹ and negative, respectively. Thus, according to adsorption energy, the adsorption of MB by purified RP was a spontaneous process, which was mainly driven by entropy increasing. Compared with neutral dye, the purified RP had higher adsorption ability for the cationic dye and anionic dye. As the purified RP dose increased from 30 to 150 mg, the adsorption capacity of purified RP increased. However, as the MB concentration and pH increased, the adsorption capacity of purified RP decreased. The purified RP had excellent reuse ability and high temperature desorption can be applied to obtain its reuse ability. Full article

The Saima alkaline rock-hosted niobium–tantalum deposit (hereafter referred to as the Saima Deposit) is situated in the Liaodong Peninsula, which constitutes the eastern segment of the northern margin of the North China Craton. The rock types of the Saima Deposit include phonolite, nepheline syenite, and aegirine nepheline syenite, which hosts niobium–tantalum ore bodies. In this study, the primary niobium-bearing minerals identified include loparite, betafite, and fersmite. The Saima pluton is characterized as a potassium-rich, low-sodium, and peraluminous alkaline pluton. Trace element characteristics reveal that the metallization-associated syenite is enriched in large-ion lithophile elements (LILEs) such as K and Rb but is relatively depleted in high-field strength elements (HFSEs). As indicated by the rare earth element (REE) profile, the Saima pluton exhibits a high total REE content (∑REE), dominance of light REEs (LREEs), and scarcity of heavy REEs (HREEs). The Sr-Nd-Pd isotopic data suggest that aegirine nepheline syenite and nepheline syenite share consistent isotopic signatures, indicating a common origin. The Saima alkaline pluton displays elevated I_Sr values ranging from 0.70712 to 0.70832 coupled with low ε_Nd(t) values between −12.84 and −11.86 and two-stage model ages (t_DM2) from 1967 to 2047 Ma. These findings indicate that the metallogenic materials for the Saima Deposit derive from both an enriched mantle source and some crustal components. The lithium (Li) isotopic fractionation observed during the genesis of the Saima pluton could be attributed to the differential diffusion rates of ⁶Li and ⁷Li under non-equilibrium fluid–rock interactions. Full article

In recent years, due to various anthropogenic activities, such as agriculture and livestock, the presence of nitrogen-associated contaminants has been increasing in surface- and groundwater resources. Among these, the main compounds present in groundwater are ammonia, nitrite, and nitrate. However, it is sometimes difficult to assess such effects given the scarcity or lack of information and the complexity of the system. In the current study, a methodology is proposed to assess nitrate in groundwater from diffuse sources considering spatiotemporal patterns of hydrological systems using a coupled SWAT/MODFLOW/MT3DMS model. The application of the model is carried out using a simplified simulation scheme of hydrological and agricultural systems because of the limited spatial and temporal data. The study area includes the Cuitzeo Lake basin in superficial flow form and the Morelia–Querendaro aquifer in groundwater flow form. The results within the methodology are surface runoff, groundwater levels, and nitrate concentrations present in surface- and groundwater systems. The results indicate that the historical and simulated nitrate concentrations were obtained within acceptable values of the statistical parameters and, therefore, are considered adequate. Full article