Generative adversarial network (GANs) is one of the most important research avenues in field artificial intelligence, and its outstanding data generation capacity has received wide attention. In this paper, we present recent progress on GANs. First, basic theory GANs differences among different generative models years were analyzed summarized. Then, derived are classified introduced by one. Third, training tricks evaluation metrics given. Fourth, applications introduced. Finally, problem,...

Recently, the Generative Adversarial Networks (GANs) are fast becoming a key promising research direction in computational intelligence. To improve modeling ability of GANs, loss functions used to measure differences between samples generated by model and real samples, make learn towards goal. In this paper, we perform survey for analyze pros cons these functions. Firstly, basic theory its training mechanism introduced. Then, GANs summarized, including not only objective but also...

In three-dimensional video system, the texture and depth videos are jointly encoded, then Depth Image Based Rendering (DIBR) is utilized to realize view synthesis. However, compression distortion of videos, as well disocclusion problem in DIBR degrade visual quality synthesized view. To address this problem, a Two-stream Attention Network (TSAN)-based enhancement method proposed for 3D-High Efficiency Video Coding (3D-HEVC) article. First, shortcomings synthesis technique traditional...

In the three-dimensional video system, depth image-based rendering is a key technique for generating synthesized views, which provides audiences with perception and interactivity. However, inaccuracy of information leads to geometrical position errors, compression distortion texture videos degrades quality views. Although existing enhancement methods can eliminate distortions in their huge computational complexity hinders applications real-time multimedia systems. To this end, residual...

Facing the shortage of energy supply and environmental pollution, solar industry is developing rapidly all over world. However, harsh volatile factors will lead to many fault types in PV systems. Simple monitoring diagnosis technology can not realize requirement intellectualization informatization power generation In this work, we present a new detection method based on Elman Neural Network (ENN). The proposed mines implicit mapping relationship between original data establish multiple...

Unprecedented mismatch occur frequently in the photovoltaic (PV) array systems, challenging classical monitoring systems. However, study of artificial intelligence methods lacks some well-designed experiments for a systematic verification because are influenced by many aspects such as PV array, DC–DC converter, DC–AC inverter, loads well environmental variables. The objective has two folds. One is to design identical apparatuses emulate 'same' real-world solar power station that operated...

Time-series 1-D signals are ubiquitous in industrial applications for monitoring and control. However, it is lacking of efficient tools to deal with simultaneously multiple time-series signals. To this end, article, a novel theory image formation proposed that converts 2-D images takes advantages convolutional neural network feature extraction classification sequence images. A case study carried out the working conditions photovoltaic power systems. In total, 23 mapped derive six different...

With the development of multimedia presentation technology, image acquisition technology and Internet industry, long-distance communication methods have changed from previous letter, audio to current audio/video. And proportion video in work, study entertainment keeps increasing, high-definition is getting more attention. Due limits network environment storage capacity, original must be encoded efficiently transmitted stored. High Efficient Video Coding (HEVC) requires a large amount time...

Abstract Most of the semantic segmentation real‐time networks improve speed by reducing spatial resolution, leading to accuracy being significantly reduced as a result. To solve this problem, we propose feature enhancement module (FEM), extraction and fusion (FEFM). By extracting enhancing future map before image down‐sample on backbone fusing low‐level features with rich details high‐level more information. Based FEM FEFM, introduce network network. In experiment, using Cityscapes CamVid...

Human motion prediction is a research field with broad application prospects. With the development of deep learning, researchers have used advanced deep-learning algorithms in this field. This paper aims to combine GRU 1D-CNN without increasing network parameters. In paper, we use learn continuity human movement, and then one-dimensional convolution networks reduce dimensions generate predicted actions. At last, utilize weight matrix which uses simple operations get from its own data, so as...