Generative Adversarial Network Architecture Research Articles

P3 AD: Privacy-Preserved Payload Anomaly Detection for Industrial Internet of Things

Payload-based anomaly detection (PAD) model is commonly built upon a big data of normal payload samples, and hence is able to discover zero-day attacks and unknown faults without the need of any negative samples in training phase. But such detection model encounters new challenges to adapt well to the emerging industrial Internet of things (IIoT). That is, the modern industrial processes are usually running in a very high complexity, resulting the payloads much more complex and diverse. Further, the industrial data is likely too sensitive to be shared in public, and thus induces a new privacy concern. To tackle these challenges, we propose PAD, a novel privacy-preserved payload-based anomaly detection model for IIoT. The basic idea is to train a two-dimensional convolutional neural network (2D-CNN) based auto-encoder using normal payloads over a federated GAN (Generative Adversarial Network) architecture, and then to detect anomalies by an unexpected dissimilarity between the original payloads and the payloads reconstructed by the auto-encoder. By the 2D-CNN, we can model the normal payloads from both the request and response directions simultaneously, and thus have more opportunities to capture the complex and dynamic industrial behaviors that are possibly reflected in the bi-directional network communications. By the GAN, we can train a more generalized auto-encoder that is able to reconstruct more general payload samples without the need to have them in advance for model training. By the federated architecture, we can remove the need of direct sharing of normal payloads, and learn them indirectly by aggregating local models across different industrial data owners, hence ensuring the payload privacy. We have evaluated PAD using four public industrial payload datasets as well as considering four typical IIoT PAD scenarios. The detection results achieve more than 0.966 in terms of F1 score for global condition and at least 0.753 for all kinds of federated settings, proving the effectiveness of our PAD with privacy preserved.

A hybrid Cycle GAN-based lightweight road perception pipeline for road dataset generation for Urban mobility.

One of the major problems that cause continual trouble in deep learning networks is that training a large network requires massive labelled datasets. The preparation of a massive labelled dataset is a cumbersome task and requires lot of human interventions. This paper proposes a novel generator network 'Sim2Real' transfer is a recent and fast-developing field in machine learning used to bridge the gap between simulated and real data. Training with simulated datasets often converges due to its size but fails to generalize real-world applications. Simulated datasets can be used to train and test deep learning models, enables the development and evaluation of new algorithms and architectures. By simulating road dataset, researchers can generate large amounts of realistic road-traffic dataset that can be used to study and understand several problems such as vehicular object tracking and classification, traffic situation analysis etc. The main advantage of such a transfer algorithm is to use the abundance of a simulated dataset to generate huge realistic-looking datasets to solve data-intense tasks. This work presents a novel, robust sim2real algorithm that converts the labels of a semantic segmentation map to a realistic-looking street view using the Cityscapes dataset and aims to achieve robust urban mobility for smart cities. Further, the generalizability of the Cycle Generative Adversarial Network (CycleGAN) architecture was tested by using an origami robot dataset for sim2real transfer. We show that the results were found to be qualitatively satisfactory for different traffic analysis applications. In addition, road perception was done using a lightweight SVM pipeline and evaluated on the KITTI dataset. We have incorporated Cycle Consistency Loss and Identity Loss as the metrics to evaluate the performance of the proposed Cycle GAN model. We inferred that the proposed Cycle GAN model provides an Identity loss of less than 0.2 in both the Cityscapes dataset and KITTI datasets. Also, we understand that the super-pixel resolution has a good impact on the quantitative results of the proposed Cycle GAN models.

Methods of applied utilization of generative adversarial networks in graphic data processing

The paper explores an important area of artificial intelligence — Generative Adversarial Networks (GANs), which are used to create high-quality artificial data samples. GANs have undergone significant development and application in various sectors, including the processing of graphical data. The report focuses on the practical use of GANs and their architecture. It discusses the fundamental principles of GAN operation, highlights the advantages and disadvantages, including issues with training, vanishing gradients, and convergence oscillations, and describes measures to overcome these problems. It also examines current research in the field of GANs and their applications in various domains, including cybersecurity, medicine, forensics, and computer vision. Practical results from the report's authors regarding their own GAN experiments, optimization, and architecture improvements are presented. The research aims to analyze the architectural features of GANs to enhance their training process

Choosing only the best voice imitators: Top-K many-to-many voice conversion with StarGAN

Voice conversion systems have become increasingly important as the use of voice technology grows. Deep learning techniques, specifically generative adversarial networks (GANs), have enabled significant progress in the creation of synthetic media, including the field of speech synthesis. One of the most recent examples, StarGAN-VC, uses a single pair of generator and discriminator to convert voices between multiple speakers. However, the training stability of GANs can be an issue. The Top-K methodology, which trains the generator using only the best K generated samples that “fool” the discriminator, has been applied to image tasks and simple GAN architectures. In this work, we demonstrate that the Top-K methodology can improve the quality and stability of converted voices in a state-of-the-art voice conversion system like StarGAN-VC. We also explore the optimal time to implement the Top-K methodology and how to reduce the value of K during training. Through both quantitative and qualitative studies, it was found that the Top-K methodology leads to quicker convergence and better conversion quality compared to regular or vanilla training. In addition, human listeners perceived the samples generated using Top-K as more natural and were more likely to believe that they were produced by a human speaker. The results of this study demonstrate that the Top-K methodology can effectively improve the performance of deep learning-based voice conversion systems.

MIGAN: GAN for facilitating malware image synthesis with improved malware classification on novel dataset

Malware visualization is a technique wherein malware binaries are represented as grayscale or color images in order to identify and extract discriminating features for classification. This technique is effectively better than classic machine learning based malware recognition techniques that require significant domain expertise or time-consuming behavioral analysis to identify discriminating features. In this manuscript, a Generative Adversarial Network (GAN) architecture is introduced for facilitating malware image synthesis called ‘MIGAN’, that can quickly produce high-quality synthetic malware images and then classify malware samples into families. The proposed framework consists of a generator and discriminator network paired with a classification module. The novelty exists in the GAN network structure, hybrid loss function, new dataset and classification network structure. The MIGAN generated images manage to achieve better Inception Score than original malware images (2.81 vs 1.90, respectively) along with better Fréchet Inception Distance score and Kernel Inception Distance score. The synthetic malware images primarily serve two purposes: firstly, it solves the class imbalance problem in custom built and public ‘Malimg’ datasets. Secondly, since these images resemble existing malware images, it is assessed to be fairly similar to upcoming ‘zero-day’ or ‘previously unseen’malware that can be eventually discovered in the future. The two classification networks (custom classification network with traditional learning approach and pretrained Resnet50v2 network with transfer learning approach) were supplemented and trained with nearly 50,000 synthetic malware images. The proposed framework achievedpromising scores of 99.2 % Area Under the Curve (AUC), 99.3 % F1-scoreand 99.5 % Accuracy. The comprehensive evaluation and excellent results demonstrate the effectiveness of the proposed framework. This framework can also be applied to image synthesis with several other types of images.

DA-VEGAN: Differentiably Augmenting VAE-GAN for microstructure reconstruction from extremely small data sets

Microstructure reconstruction is an important and emerging field of research and an essential foundation to improving inverse computational materials engineering (ICME). Much of the recent progress in the field is made based on generative adversarial networks (GANs). Although excellent results have been achieved throughout a variety of materials, challenges remain regarding the interpretability of the model’s latent space as well as the applicability to extremely small data sets. The present work addresses these issues by introducing DA-VEGAN, a model with two central innovations. First, a β-variational autoencoder is incorporated into a hybrid GAN architecture that allows to penalize strong nonlinearities in the latent space by an additional parameter, β. Secondly, a custom differentiable data augmentation scheme is developed specifically for this architecture. The differentiability allows the model to learn from extremely small data sets without mode collapse or deteriorated sample quality. An extensive validation on a variety of structures demonstrates the potential of the method and future directions of investigation are discussed.

Lensless Image Restoration Based on Multi-Stage Deep Neural Networks and Pix2pix Architecture

Lensless imaging represents a significant advancement in imaging technology, offering unique benefits over traditional optical systems due to its compact form factor, ideal for applications within the Internet of Things (IoT) ecosystem. Despite its potential, the intensive computational requirements of current lensless imaging reconstruction algorithms pose a challenge, often exceeding the resource constraints typical of IoT devices. To meet this challenge, a novel approach is introduced, merging multi-level image restoration with the pix2pix generative adversarial network architecture within the lensless imaging sphere. Building on the foundation provided by U-Net, a Multi-level Attention-based Lensless Image Restoration Network (MARN) is introduced to further augment the generator’s capabilities. In this methodology, images reconstructed through Tikhonov regularization are perceived as degraded images, forming the foundation for further refinement via the Pix2pix network. This process is enhanced by incorporating an attention-focused mechanism in the encoder--decoder structure and by implementing stage-wise supervised training within the deep convolutional network, contributing markedly to the improvement of the final image quality. Through detailed comparative evaluations, the superiority of the introduced method is affirmed, outperforming existing techniques and underscoring its suitability for addressing the computational challenges in lensless imaging within IoT environments. This method can produce excellent lensless image reconstructions when sufficient computational resources are available, and it consistently delivers optimal results across varying computational resource constraints. This algorithm enhances the applicability of lensless imaging in applications such as the Internet of Things, providing higher-quality image acquisition and processing capabilities for these domains.

RAGAN: A Generative Adversarial Network for risk-aware trajectory prediction in multi-ship encounter situations

Ship trajectory prediction plays a vital role in ensuring safety in dense waterway traffic. However, conventional prediction methods often disregard the intricate spatial–temporal interactions and the inherent collision avoidance maneuvers that occur during multiple ship encounters, leading to inadequate accuracy in interacted trajectory predictions. To address this challenge, we propose a risk-aware trajectory prediction framework based on Generative Adversarial Network (RAGAN) architectures. RAGAN leverage GANs to learn latent collision avoidance interaction patterns in multi-ship encounters. Through adversarial training, RAGAN continuously enhances its capability to generate precise and safely acceptable ship trajectories that faithfully represent the actual collision avoidance strategies adopted by ship officers. Furthermore, by incorporating an interacting gate function, RAGAN effectively captures the evolving risks and influences within encounter situations, enabling dynamic representation of ship interactions and their profound impact on the development of encounters. Evaluation using real-world ship trajectory data demonstrates the superior performance of RAGAN compared to the baseline method across various evaluation metrics. Through case studies, RAGAN exhibits the ability to learn navigation rules beyond existing regulations, such as the COLREGs, and provides valuable insights for enhancing safety in maritime operations.

Generative Adversarial Network with Convolutional Wavelet Packet Transforms for Automated Speaker Recognition and Classification

Speech is an effective mode of communication that always conveys abundant and pertinent information, such as the gender, accent, and other distinguishing characteristics of the speaker. These distinctive characteristics allow researchers to identify human voices using artificial intelligence (AI) techniques, which are useful for forensic voice verification, security and surveillance, electronic voice eavesdropping, mobile banking, and mobile purchasing. Deep learning (DL) and other advances in hardware have piqued the interest of researchers studying automatic speaker identification (SI). In recent years, Generative Adversarial Networks (GANs) have demonstrated exceptional ability in producing synthetic data and improving the performance of several machine learning tasks. The capacity of Convolutional Wavelet Packet Transform (CWPT) and Generative Adversarial Networks are combined in this paper to propose a novel way of enhancing the accuracy and robustness of Speaker Recognition and Classification systems. Audio signals are dissected using the Convolutional Wavelet Packet Transform into a multi-resolution, time-frequency representation that faithfully preserves local and global characteristics. The improved audio features better precisely describe speech traits and handle pitch, tone, and pronunciation variations that are frequent in speaker recognition tasks. Using GANs to create synthetic speech samples, our suggested method GAN-CWPT enriches the training data and broadens the dataset's diversity. The generator and discriminator components of the GAN architecture have been tweaked to produce realistic speech samples with attributes quite similar to genuine speaker utterances. The new dataset enhances the Speaker Recognition and Classification system's robustness and generalization, even in environments with little training data. We conduct extensive tests on standard speaker recognition datasets to determine how well our method works. The findings demonstrate that, compared to conventional methods, the GAN-CWPTs combination significantly improves speaker recognition, classification accuracy, and efficiency. Additionally, the suggested model GAN-CWPT exhibits stronger generalization on unknown speakers and excels even with loud and poor audio inputs.

Blind Attention Geometric Restraint Neural Network for Single Image Dynamic/Defocus Deblurring.

Based on the information loss analysis of the blur accumulation model, a novel single-image deblurring method is proposed. We apply the recurrent neural network architecture to capture the attention perception map and the generative adversarial network (GAN) architecture to yield the deblurring image. Considering that the attention mechanism has to make hard decisions about specific parts of the input image to be focused on since blurry regions are not given, we propose a new adaptive attention disentanglement model based on the variation blind source separation, which provides the global geometric restraint to reduce the large solution space, so that the generator can realistically restore details on blurry regions, and the discriminator can accurately assess the content consistency of the restored regions. Since we combine blind source separation, attention geometric restraint with GANs, we name the proposed method BAGdeblur. Extensive evaluations on quantitative and qualitative experiments show that the proposed method achieves the state-of-the-art performance on both synthetic datasets and real-world blurry images.

GENERATIVE ADVERSARIAL NETWORKS FOR IMAGE SYNTHESIS AND STYLE TRANSFER IN VIDEOS

In computer vision and artistic expression, the synthesis of visually compelling images and the transfer of artistic styles onto videos have gained significant attention. This research addresses the challenges in achieving realistic image synthesis and style transfer in the dynamic context of videos. Existing methods often struggle to maintain temporal coherence and fail to capture intricate details, prompting the need for innovative approaches. The conventional methods for image synthesis and style transfer in videos encounter difficulties in preserving the natural flow of motion and consistency across frames. This research aims to bridge this gap by leveraging the power of Generative Adversarial Networks (GANs) to enhance the quality and temporal coherence of synthesized images in video sequences. While GANs have demonstrated success in image generation, their application to video synthesis and style transfer remains an underexplored domain. The research seeks to address this gap by proposing a novel methodology that optimizes GANs for video-challenges, aiming for realistic, high-quality, and temporally consistent results. Our approach involves the development of a specialized GAN architecture tailored for video synthesis, incorporating temporal-aware modules to ensure smooth transitions between frames. Additionally, a style transfer mechanism is integrated, enabling the transfer of artistic styles onto videos seamlessly. The model is trained on diverse datasets to enhance its generalization capabilities. Experimental results showcase the efficacy of the proposed methodology in generating lifelike images and seamlessly transferring styles across video frames. Comparative analyses demonstrate the superiority of our approach over existing methods, highlighting its ability to address the temporal challenges inherent in video synthesis and style transfer.

Generative Adversarial Networks (GANs) for Audio-Visual Speech Recognition in Artificial Intelligence IoT

This paper proposes a novel multimodal generative adversarial network AVSR (multimodal AVSR GAN) architecture, to improve both the energy efficiency and the AVSR classification accuracy of artificial intelligence Internet of things (IoT) applications. The audio-visual speech recognition (AVSR) modality is a classical multimodal modality, which is commonly used in IoT and embedded systems. Examples of suitable IoT applications include in-cabin speech recognition systems for driving systems, AVSR in augmented reality environments, and interactive applications such as virtual aquariums. The application of multimodal sensor data for IoT applications requires efficient information processing, to meet the hardware constraints of IoT devices. The proposed multimodal AVSR GAN architecture is composed of a discriminator and a generator, each of which is a two-stream network, corresponding to the audio stream information and the visual stream information, respectively. To validate this approach, we used augmented data from well-known datasets (LRS2-Lip Reading Sentences 2 and LRS3) in the training process, and testing was performed using the original data. The research and experimental results showed that the proposed multimodal AVSR GAN architecture improved the AVSR classification accuracy. Furthermore, in this study, we discuss the domain of GANs and provide a concise summary of the proposed GANs.

Utilizing Generative Adversarial Networks for Stable Structure Generation in Angry Birds

This paper investigates the suitability of using Generative Adversarial Networks (GANs) to generate stable structures for the physics-based puzzle game Angry Birds. While previous applications of GANs for level generation have been mostly limited to tile-based representations, this paper explores their suitability for creating stable structures made from multiple smaller blocks. This includes a detailed encoding/decoding process for converting between Angry Birds level descriptions and a suitable grid-based representation, as well as utilizing state-of-the-art GAN architectures and training methods to produce new structure designs. Our results show that GANs can be successfully applied to generate a varied range of complex and stable Angry Birds structures.

Multivariate Emulation of Kilometer-Scale Numerical Weather Predictions with Generative Adversarial Networks: A Proof of Concept

Abstract Emulating numerical weather prediction (NWP) model outputs is important to compute large datasets of weather fields in an efficient way. The purpose of the present paper is to investigate the ability of generative adversarial networks (GANs) to emulate distributions of multivariate outputs (10-m wind and 2-m temperature) of a kilometer-scale NWP model. For that purpose, a residual GAN architecture, regularized with spectral normalization, is trained against a kilometer-scale dataset from the AROME Ensemble Prediction System (AROME-EPS). A wide range of metrics is used for quality assessment, including pixelwise and multiscale Earth-mover distances, spectral analysis, and correlation length scales. The use of wavelet-based scattering coefficients as meaningful metrics is also presented. The GAN generates samples with good distribution recovery and good skill in average spectrum reconstruction. Important local weather patterns are reproduced with a high level of detail, while the joint generation of multivariate samples matches the underlying AROME-EPS distribution. The different metrics introduced describe the GAN’s behavior in a complementary manner, highlighting the need to go beyond spectral analysis in generation quality assessment. An ablation study then shows that removing variables from the generation process is globally beneficial, pointing at the GAN limitations to leverage cross-variable correlations. The role of absolute positional bias in the training process is also characterized, explaining both accelerated learning and quality-diversity trade-off in the multivariate emulation. These results open perspectives about the use of GAN to enrich NWP ensemble approaches, provided that the aforementioned positional bias is properly controlled.

Four-channel generative adversarial networks can predict the distribution of reef-associated fish in the South and East China Seas

Reef-associated fish (RAF) are socioeconomically and ecologically important. Herein, we used a four-channel generative adversarial network (GAN) architecture with a retrospective cycle feature to model the distribution of RAF in the East China Sea (ECS) and South China Sea (SCS). Our approach entailed the use of RAF suitability maps and key environmental layers, including sea body temperature, chlorophyll-α concentration, and salinity, as input data. These environmental factors have been demonstrated to significantly influence the distribution of RAF. Our results show that four-channel GANs performed better than one-channel GANs in predicting the distribution of RAF. Furthermore, the proposed four-channel retrospective-cycle GAN (4CRCGAN) model outperformed the commonly used four-channel convolutional GAN (4CCGAN), emphasizing the importance of considering temporal dynamics and feedback loops in predicting the distribution of species. However, it is worth noting that the predictive performance for the distribution of RAF in summer and winter was relatively poor, implying that other environmental and biological factors also exert nonnegligible influences on fish movement and abundance; this underscores the necessity of comprehensively understanding the intricate and dynamic nature of species-environment interactions within marine ecosystems. Our study also revealed that the proposed 4CRCGAN model performed well over multiple steps, which is reflective of its potential to provide valuable insights for decision-makers and stakeholders in the fisheries management sector. Overall, our findings offer critical insights for the development of effective and sustainable management strategies that balance the conservation and sustainable use of these valuable marine resources.

Adaptive Filter in Single Image SRGAN

Abstract: SRGAN is a generative adversarial network for super-resolution of a single picture. In this paper an adaptive filter in single image SRGAN which is a Super-Resolution Generative Adversarial Network (SRGAN), is a deep learning-based approach that is used to generate high-resolution images from low-resolution ones. The SRGAN model is based on the Generative Adversarial Network (GAN) architecture, which consists of two deep neural networks: a generator network and a discriminator network. The generator network takes a low-resolution image as input and tries to generate a high-resolution image that is similar to the original high-resolution image. The discriminator network, on the other hand, tries to distinguish between the high-resolution images generated by the generator network and the original high-resolution images. The SRGAN model has been shown to be very effective in generating high-quality, realistic images from low-resolution inputs, and it has been applied in various applications, including image and video upscaling, medical imaging, and satellite imaging

Prior-guided generative adversarial network for mammogram synthesis

Deep Learning is vital in medical imaging solutions and clinical applications. However, multiple reasons, such as data scarcity and imbalance in the medical image dataset, cause performance issues in various deep learning models. Thus, generating synthetic medical data close to real images is an immediate need of time. The Mammographic Image Analysis Society (MIAS) dataset, the standard dataset for studying breast anomalies, suffers from a class imbalance problem as very few images correspond to malignant and benign cases are there compared to normal cases. Methods:This paper proposes a data augmentation model based on Generative Adversarial Network (GAN) architecture to generate synthetic mammograms of different cases, such as normal, benign, and malignant. The proposed method’s novelty lies in its capability of generating multiple variants of a class-labelled mammogram, which is more realistic in conserving the adherent breast tissue characteristics. These synthetic mammograms are a useful solution for resolving class imbalance problems. Results:Generated samples of each class are added to the MIAS dataset to address the class imbalance problem. This study also demonstrates that the augmented dataset with cGAN-generated images has enhanced the 2-class- and 3-class breast cancer classification performance by 3.9%. The proposed cGAN model is efficient in capturing the features of abnormalities. The 90% of class-labelled mammograms generated by cGAN belong to the same class in the case of real medical data. Conclusions:The proposed work exhibits the generation of synthetic mammograms using a GAN-based approach. Augmentation of the training dataset with GAN-generated images significantly enhances breast image classification.

A deep adversarial approach for the generation of synthetic titanium alloy microstructures with limited training data

Microstructures play a central role in determining the mechanical and functional properties of a material. An important aspect in computational materials science is to reliably predict the key microstructural features, and to utilise them in bridging the processing and properties of new materials. The existing microstructure characterization and reconstruction (MCR) techniques have inherent limitations in terms of the lack of design variables and information loss due to various assumptions. In previous studies, Generative Adversarial Network (GAN) models were trained using more than 10,000 image datasets, however without performance analysis using quantitative morphometric and statistical measures. In this perspective, the present work demonstrates the capability of GAN architectures to learn the mapping between random latent vectors and synthetic microstructural images, even in a limited data regime (1225 images). Three different architectures Deep Convolutional GAN (DCGAN) ,Wasserstein GAN- Gradient Penalty (WGAN-GP), StyleGAN2- Adaptive Discriminator Augmentation (ADA) were explored, together with comprehensive statistical and morphological analysis, while training on a publicly accessible Ti-6Al-4V (Ti64) alloy microstructure dataset. The StyleGAN2-ADA outperforms the other two GAN models to generate realistic synthetic microstructures of higher resolution, with good qualitative similarity to original images. The analysis of performance metrics, like Frechet Inception Distance (FID), Kernel Inception Distance (KID), and Inception Score (IS) scores, reveal that the generated image distribution is statistically close to the original distribution of microstructural features. The morphometric parameters, including α/β phase fractions, local α/β boundary thickness, and orientation of lamellar morphology, were also used to compare original and synthetic images quantitatively. More importantly, two-point correlation and t-stochastic neighbour embedding (t-SNE) illustrate the statistical similarity between the original and synthetic microstructures. Taken together, the present work establishes the capability of generative models like GAN in generating representative microstructures of Titanium alloy in a statistically reliable manner. Such an approach, when adopted, will accelerate the field of microstructure fingerprinting.

Synthetic dual-energy CT reconstruction from single-energy CT Using artificial intelligence.

To develop and assess the utility of synthetic dual-energy CT (sDECT) images generated from single-energy CT (SECT) using two state-of-the-art generative adversarial network (GAN) architectures for artificial intelligence-based image translation. In this retrospective study, 734 patients (389F; 62.8 years ± 14.9) who underwent enhanced DECT of the chest, abdomen, and pelvis between January 2018 and June 2019 were included. Using 70-keV as the input images (n = 141,009) and 50-keV, iodine, and virtual unenhanced (VUE) images as outputs, separate models were trained using Pix2PixHD and CycleGAN. Model performance on the test set (n = 17,839) was evaluated using mean squared error, structural similarity index, and peak signal-to-noise ratio. To objectively test the utility of these models, synthetic iodine material density and 50-keV images were generated from SECT images of 16 patients with gastrointestinal bleeding performed at another institution. The conspicuity of gastrointestinal bleeding using sDECT was compared to portal venous phase SECT. Synthetic VUE images were generated from 37 patients who underwent a CT urogram at another institution and model performance was compared to true unenhanced images. sDECT from both Pix2PixHD and CycleGAN were qualitatively indistinguishable from true DECT by a board-certified radiologist (avg accuracy 64.5%). Pix2PixHD had better quantitative performance compared to CycleGAN (e.g., structural similarity index for iodine: 87% vs. 46%, p-value < 0.001). sDECT using Pix2PixHD showed increased bleeding conspicuity for gastrointestinal bleeding and better removal of iodine on synthetic VUE compared to CycleGAN. sDECT from SECT using Pix2PixHD may afford some of the advantages of DECT.

Deep neural architecture for natural language image synthesis for Tamil text using BASEGAN and hybrid super resolution GAN (HSRGAN)

Tamil is a language that has the most extended history and is a conventional language of India. It has antique origins and a distinct tradition. A study reveals that at the beginning of the twenty-first century, more than 66 million people spoke Tamil. In the present time, image synthesis from text emerged as a promising advancement in computer vision applications. The research work done so far in intelligent systems is trained in universal language but still has not achieved the desired development level in regional languages. Regional languages have a greater scope for developing applications and will enhance more research areas to be explored, ruling out the barrier. The current work using Auto Encoders failed at the point of providing vivid information along with essential descriptions of the synthesised images. The work aims to generate embedding vectors using a language model headed by image synthesis using GAN (Generative Adversarial Network) architecture. The proposed method is divided into two stages: designing a language model TBERTBASECASE model for generating embedding vectors. Synthesising images using Generative Adversarial Network called BASEGAN, the resolution has been improved through two-stage architecture named HYBRID SUPER RESOLUTION GAN. The work uses Oxford-102 and CUB-200 datasets. The framework efficiency has been measured using F1 Score, Fréchet inception distance (FID), and Inception Score (IS). Language and image synthesis architecture proposed can bridge the gap between the research ideas in regional languages.