Channel is one of the most important parts of a communication system as the medium of the propagation of electromagnetic waves. Being aware of how the channel affects the propagation waves is essential for the design, optimization, and performance analysis of a communication system. Along with conventional modeling schemes, in this paper, we present a novel propagation channel model. The proposed modeling strategy considers the 2-dimensional time-frequency response of the channel as an image. It models the distribution of these channel images using Deep Convolutional Generative Adversarial Networks (DCGANs). In addition, for the measurements with different user speeds, the user speed is considered as an auxiliary parameter for the model. StarGAN is used as an image-to-image translation technique to change the generated channel images with respect to the desired user speed. The performance of the proposed model is evaluated using a few existing evaluation metrics. Furthermore, as modeling the 2D time- frequency response is more general than the modeling of the channel only in time, the conventional metrics for evaluation of the channel models are not sufficient; therefore, a new metric is introduced in this paper. This metric is based on the Cepstral Distance Measure (CDM) between the mean autocorrelation of the generated samples and measurement data. Using this metric, the generated channels show significant statistical similarity to the measurement data.

Generative models try to obtain a probability distribution that is similar to that of observed data. Two different solutions have been proposed in this regard in recent years: one is to minimize the divergence (distance) between the two distributions by maximizing the variational lower bound, and the other is to implicitly reduce the distance between the two distributions through Adversarial processes. One of the problems in Generative Adversarial Networks (GANs) is the mode collapse. Mode collapse is a phenomenon in which, for various inputs, the Generative model generates low variety or similar images. This paper tries to provide a solution to the mode collapse problem proposing a novel method called variational Generative Adversarial Networks (VGANs). This method exploits variational autoencoders to initialize GANs. In other words, in addition to maximizing the variational lower bound, it also implicitly reduces the distance between the two distributions. Experimental results show that this method can deal with the mode collapse problem better than the state-of-the-art. Moreover, in the qualitative analysis, according to a survey of 136 people on the authenticity of the generated images, the proposed method can generate images more similar to real ones.

Translating face sketches to photo-realistic faces is an interesting and essential task in many applications like law enforcement and the digital entertainment industry. One of the most important challenges of this task is the inherent differences between the sketch and the real image such as the lack of color and details of the skin tissue in the sketch. With the advent of Adversarial Generative models, an increasing number of methods have been proposed for sketch-to-image synthesis. However, these models still suffer from limitations such as the large number of paired data required for training, the low resolution of the produced images, or the unrealistic appearance of the generated images. In this paper, we propose a method for converting an input facial sketch to a colorful photo without the need for any paired dataset. To do so, we use a pre-trained face photo generating model to synthesize high-quality natural face photos and employ an optimization procedure to keep high fidelity to the input sketch. We train a network to map the facial features extracted from the input sketch to a vector in the latent space of the face-generating model. Also, we study different optimization criteria and compare the results of the proposed model with those of the state-of-the-art models quantitatively and qualitatively. The proposed model achieved 0.655 in the SSIM index and 97.59% rank-1 face recognition rate with a higher quality of the produced images.

Automatic colorization of gray scale images poses a unique challenge in Information Retrieval. The goal of this field is to colorize images which have lost some color channels (such as the RGB channels or the AB channels in the LAB color space) while only having the brightness channel available, which is usually the case in a vast array of old photos and portraits. Having the ability to colorize such images would give us a multitude of possibilities ranging from colorizing old and historic images to providing alternate colorizations for real images or artistic creations. Be that as it may, the progress in this field is trivial compared to what the professionals are able to do using special-purpose applications such as Photoshop or GIMP. On the other hand, losing the information stored in color channels and having only access to the primary brightness channel, makes this problem a unique challenge, since the main aim of automatic colorization is not to find the image’ s “ real” color but to colorize it in such a way that makes it “ seem real” as the original color information is lost forever and the only way to colorize it, is to provide a somewhat “ proper” estimation. In this research we propose a model to automatically colorize gray human portraits. We start by reviewing the methods used for the task of image colorization and provide an explanation as to why most of them collapse to a situation known as “ Averaging” . To counteract this effect, we design our end-to-end model with two separate deep neural Networks forming a Generative Adversarial Network (GAN), one to colorize the images and the other to evaluate the colorization of the first network and guide it towards the proper distribution. The results show improvements over other proposed methods in this field especially in the case of colorizing human portraits along faster train times. This method not only works on real human portraits but also on non-human and artistic portraits that can be leveraged to colorize hand-drawn images some of which may take minutes up to hours by hand.

The retina may be affected by many diseases such as Age-related Macular Degeneration (AMD), Diabetic Macular Disease (DME), and Choroidal Neovascularization (CNV). To diagnose these diseases, one way is to analyze retinal Optical Coherence Tomography (OCT) images using image processing algorithms. In this paper, a novel architecture based on Generative Adversarial Networks (GANs) and Convolutional Neural Networks (CNNs) is proposed to classify OCT images with insufficient training samples. The proposed method generates OCT images with pseudo labels from the distribution of original OCT images. OCT samples with real and pseudo labels are presented to a CNN classifier which leads to a model which is robust against insufficient samples. UCSD dataset has been used to evaluate the proposed method. Results indicate that the proposed method is comparable to the state-of-the-art methods in the terms of Precision, Sensitivity, Specificity, and F-Measure. Source code of this paper is available online at Github (https://github.com/mohamad-sw/oct-image-classification-using-gans).

A significant amount of Iranian hydrocarbon resources is produced from fractured reservoirs with tight rock matrices. The structure of pores in these reservoirs is so complex. Very tiny pores and throats in nanometer sizes are responsible for reserving hydrocarbons. By understanding the structure of porous media and examining fluid flow inside these nanometer pores, we can better understand the porous media›s behaviour on larger scales. Investigating fluid flow in reservoir rocks requires three-dimensional structures with appropriate accuracy. However, using conventional methods to reconstruct a porous medium is expensive. On the other hand, as these structures become more complex, the ability of these methods to reconstruct pore network models decreases significantly. In recent years, with the advance in computer science, especially artificial intelligence, a new gate has been opened for reconstructing complex structures such as tight reservoir rocks. By implementing machine learning methods, three-dimensional pore-scale models can be created with high accuracy. The petrophysical properties of rocks can be calculated from them. One of these methods is the Generative Adversarial network (GAN), which has proven to reconstruct the pore structure of rocks. This study uses a GAN with convolutional layers to reconstruct the images obtained from FIB-SEM of a tight reservoir rock at the pore scale. Different realizations of the pore space are reconstructed by the trained GAN. The porosity and permeability of the reconstructed images are very close to the properties in the actual FIB-SEM image and have a deviation of 1. 07% and 5. 24%, respectively. It can be seen that GANs have a high capacity in rock reconstruction at the pore scale, especially for tight reservoirs.

Generating dynamic videos from static images and accurately modeling object motion within scenes are fundamental challenges in computer vision, with broad applications in video enhancement, photo animation, and visual scene understanding. This paper proposes a novel hybrid framework that combines convolutional neural Networks (CNNs), recurrent neural Networks (RNNs) with long short-term memory (LSTM) units, and Generative Adversarial Networks (GANs) to synthesize temporally consistent and spatially realistic video sequences from still images. The architecture incorporates splicing techniques, the Lucas-Kanade motion estimation algorithm, and a loop feedback mechanism to address key limitations of existing approaches, including motion instability, temporal noise, and degraded video quality over time. CNNs extract spatial features, LSTMs model temporal dynamics, and GANs enhance visual realism through Adversarial training. Experimental results on the KTH dataset, comprising 600 videos of fundamental human actions, demonstrate that the proposed method achieves substantial improvements over baseline models, reaching a peak PSNR of 35.8 and SSIM of 0.96—representing a 20% performance gain. The model successfully generates high-quality, 10-second videos at a resolution of 720×1280 pixels with significantly reduced noise, confirming the effectiveness of the integrated splicing and feedback strategy for stable and coherent video generation.

This research is related to the use of deep learning tools and image processing technology in the automatic generation of images from text. Previous researches have used one sentence to produce images. In this research, a memory-based hierarchical model is presented that uses three different descriptions that are presented in the form of sentences to produce and improve the image. The proposed scheme focuses on using more information to produce high-resolution images, using competitive productive Networks. Implementing programs related to this field require massive processing resources. Therefore, the proposed method was implemented and tested on a cluster with 25 GPUs using the hardware platform of the University of Copenhagen. The experiments were performed on CUB-200 and ids-ade datasets. The experimental results show that the proposed model can produce higher quality images than the two basic models StackGAN and AttGAN.

The diagnosis of cancer is mainly performed by visual analysis of pathologists through examining the morphology of the tissue slices under a microscope. If the microscopic image of a specimen is not stained, it will look colorless and without texture. Therefore, chemical staining is required to create adequate contrast and help identify specific tissue components. During tissue preparation due to differences in chemicals, scanners, and types of illness, similar tissues are usually varied significantly in appearance. This diversity in staining, in addition to interpretive disparity among pathologists, is one of the main challenges in designing robust and flexible systems for automated analysis. Various strategies for stain normalization have been proposed as a pre-processing step in the pipeline of the automated systems. The pix2pix methodwhich is derived from the conditional Generative Adversarial Networks (cGAN) is one of the powerful methods for solving image-to-image translation problems. The main innovation of this paper is to present a new powerful method for the stain normalization of histopathology images using the Pix2Pix method, which is implemented and evaluated on the Mitos-Atypia-14 dataset.In the proposed method, grayscale images are given as input to the network, and then the system learns to restain the texture of the input image in a specific coloring style by preserving the structure and corresponding histopathological pattern. This method, compared to previous methods that relied on a reference image, instead uses the distribution of all images in the learning phase. The proposed method has achieved significant resultsboth in quantitative and qualitative evaluations comparing to some well-known methods in the literature.Moreover, as another innovation, the proposed method tested in a clinical use-case, namely breast cancer tumor classification,using the PatchCamelyon datasetand itshowsa 5% increase in the AUC parameter.