An analysis of generative adversarial networks and variants for image synthesis on MNIST dataset

Generative Adversarial Networks (GANs) are most popular generative frameworks that have achieved compelling performance. They follow an adversarial approach where two deep models generator and discriminator compete with each other. They have been used for many applications especially for image synthesis because of their capability to generate high quality images. In past few years, different variants of GAN have proposed and they produced high quality results for image generation. This paper conducts an analysis of working and architecture of GAN and its popular variants for image generation in detail. In addition, we summarize and compare these models according to different parameters such as architecture, training method, learning type, benefits and performance metrics. Finally, we apply all these methods on a benchmark MNIST dataset, which contains handwritten digits and compare qualitative and quantitative results. The evaluation is based on quality of generated images, classification accuracy, discriminator loss, generator loss and computational time of these models. The aim of this study is to provide a comprehensive information about GAN and its various models in the field of image synthesis. Our main contribution in this work is critical comparison of popular GAN variants for image generation on MNIST dataset. Moreover, this paper gives insights regarding existing limitations and challenges faced by GAN and discusses associated future research work.

     

基于生成对抗网络的医学图像超分辨率重建

针对医学图像分辨率低导致视觉效果差的问题,提出一种基于生成对抗网络的医学图像超分辨率重建方法.使用生成对抗网络架构,由生成器重建高分辨率图像,再将生成器生成的高分辨率图像送入判别器判断真伪.通过实验验证了该方法的有效性,在视觉效果和数值结果上都有所提高.     

A generative adversarial network for image denoising

Multimedia Tools and Applications - Recent studies have shown that the performance of image denoising methods can be improved significantly by using deep convolutional neural networks(CNN). The...     

GANFuse: a novel multi-exposure image fusion method based on generative adversarial networks

Neural Computing and Applications - In this paper, a novel multi-exposure image fusion method based on generative adversarial networks (termed as GANFuse) is presented. Conventional multi-exposure...     

Invisible steganography via generative adversarial networks

Multimedia Tools and Applications - Nowadays, there are plenty of works introducing convolutional neural networks (CNNs) to the steganalysis and exceeding conventional steganalysis algorithms....     

MDVA-GAN: multi-domain visual attribution generative adversarial networks

Neural Computing and Applications - Some pixels of an input image have thick information and give insights about a particular category during classification decisions. Visualization of these pixels...     

Scalable balanced training of conditional generative adversarial neural networks on image data

The Journal of Supercomputing - We propose a distributed approach to train deep convolutional generative adversarial neural network (DC-CGANs) models. Our method reduces the imbalance between...     

Boosting attention fusion generative adversarial network for image denoising

Neural Computing and Applications - Boosting has received considerable attention to improve the overall performance of model in multiple tasks by cascading many steerable sub-modules. In this...     

Triple discriminator generative adversarial network for zero-shot image classification

One key challenge in zero-shot classification(ZSC)is the exploration of knowledge hidden in unseen classes.Generative methods such as generative adversarial net...     

Finger vein image inpainting using neighbor binary-wasserstein generative adversarial networks (NB-WGAN)

Traditional inpainting methods obtain poor performance for finger vein images with blurred texture. In this paper, a finger vein image inpainting method using Neighbor Binary-Wasserstein Generative Adversarial Networks (NB-WGAN) is proposed. Firstly, the proposed algorithm uses texture loss, reconstruction loss, and adversarial loss to constrain the network, which protects the texture in the inpainting process. Secondly, the proposed NB-WGAN is designed with a coarse-to-precise generator network and a discriminator network composed of two Wasserstein Generative Adversarial Networks with Gradient Penalty (WGAN-GP). The cascade of a coarse generator network and a precise generator network based on Poisson fusion can obtain richer information and get natural boundary connection. The discriminator consists of a global WGAN-GP and a local WGAN-GP, which enforces consistency between the entire image and the repaired area. Thirdly, a training dataset is designed by analyzing the locations and sizes of the damaged finger vein images in practical applications (i.e., physical oil dirt, physical finger molting, etc). Experimental results show that the performance of the proposed algorithm is better than traditional inpainting methods including Curvature Driven Diffusions algorithm without texture constraints, a traditional inpainting algorithm with Gabor texture constraints, and a WGAN inpainting algorithm based on attention mechanism without texture constraints.

     

A framework for personalized recommendation with conditional generative adversarial networks

Knowledge and Information Systems - Recommender systems suffer from interaction data sparsity in reality. Recently, generative adversarial network-based recommender systems have shown the potential...     

Data augmentation for handwritten digit recognition using generative adversarial networks

Multimedia Tools and Applications - Supervised learning techniques require labeled examples that can be time consuming to obtain. In particular, deep learning approaches, where all the feature...     

Dummy trajectory generation scheme based on generative adversarial networks

Neural Computing and Applications - Dummy trajectory is widely used to protect the privacy of mobile users’ locations. However, two main challenges remain: (1) Map background information has...     

基于条件生成对抗网络的视频显著性目标检测

针对传统的显著性检测方法存在着流程复杂,计算成本高,特征学习不足等问题,受生成对抗网络以及弹性网络的启发,提出一种基于条件生成对抗网络(c GAN)与L1,L2范式联合正则的视频显著性目标检测方法。方法需训练2个模型:生成器和判别器。生成器尽可能生成与真实值一样的显著图来迷惑判别器,使其难以辨别预测的显著图的真实性。判别器则尽可能准确地区分"假"显著图。实验表明:所提方法在两个公开视频数据集上的检测效果都超过了当前主流方法,且算法流程简单,运算效率更高。     

Designing mm-wave electromagnetic engineered surfaces using generative adversarial networks

Neural Computing and Applications - In this paper, we investigate the capability of generative adversarial networks, including conditional and conditional convolutional generative adversarial...     

Self-attention generative adversarial networks applied to conditional music generation

The task of audio and music generation in the waveform domain has become possible due to recent advances in deep learning. Generative Adversarial Networks (GANs) are a type of generative model that has achieved success in areas such as image, video and audio generation. However, realistic audio generation with GANs is still a challenge, thanks to the specific characteristics inherent to this kind of data. In this paper we propose a GAN model that employs the self-attention mechanism and produces small chunks of music conditioned by instrument. We compare our model to a baseline and run ablation studies in order to demonstrate its superiority. We also suggest some applications of the model, particularly in the area of computer assisted composition.

     

SCGAN:合成单类别表格数据的生成对抗网络

为了提升合成表格数据的质量,提出一种简单的方法生成每个类的数据,使用度量损失控制每一类结构化数据的生成,将此方法命名为SCGAN.文章用此方法在二分类问题上进行了尝试.使用三种不同的度量损失在三个真实的数据集上训练生成对抗网络:逐次对每一类数据进行合成,利用合成数据训练分类器模型,使用gmean来评估模型的性能.结果表...     

AdvSGAN: Adversarial image Steganography with adversarial networks

Multimedia Tools and Applications - Steganalysers based on deep learning achieve state-of-the-art performance. However, due to the difficulty of capturing the distribution of the high-dimensional...     

Large-area damage image restoration algorithm based on generative adversarial network

Neural Computing and Applications - Given that the traditional image restoration algorithm cannot generate high-quality false images and the restoration accuracy for the large-area damaged images...     

Depth map prediction from a single image with generative adversarial nets

Multimedia Tools and Applications - A depth map is a fundamental component of 3D construction. Depth map prediction from a single image is a challenging task in computer vision. In this paper, we...