Convolutional neural network (CNN) has achieved remarkable success in the field of fundus images due to its powerful feature learning ability. Computer-aided diagnosis can obtain information with reference value for doctors clinical or screening through proper processing and analysis images. However, most previous studies have focused on detection a certain disease, simultaneous multiple diseases still faces great challenges. We propose multi-label classification ensemble model based CNN...

Abstract Zero-Reference Deep Curve Estimation (Zero-DCE) is currently one of the most popular low-light image enhancement methods. Through extensive experimentation, we observe that: (i) excellent performance Zero-DCE depends on training data with multiple exposure levels, (ii) it cannot effectively handle uneven light, extremely low or overexposed images in natural environments. Therefore, propose an improved zero-reference dual-illumination deep curve estimation method for named...

Abstract Although existing object detectors achieve encouraging performance of detection and localisation under real ideal conditions, the in adverse weather conditions (snowy) is very poor not enough to cope with task conditions. Existing methods do deal well effect snow on identity features or usually ignore even discard potential information that can help improve performance. To this end, authors propose a novel improved end‐to‐end network joint image restoration. Specifically, order...

Most real-time semantic segmentation networks use shallow architectures to achieve fast inference speeds. This approach, however, limits a network’s receptive field. Concurrently, feature information extraction is restricted single scale, which reduces the ability generalize and maintain robustness. Furthermore, loss of image spatial details negatively impacts accuracy. To address these limitations, this paper proposes Multiscale Context Pyramid Pooling Spatial Detail Enhancement Network...

On the basis of feature points pairing, a scale‐invariant matching method is proposed in this study. The distance between two features used to compute pairs' support region size, which different from methods using detectors provide information find region. Moreover, achieve rotation invariance, sub‐region division based on intensity order introduced. For comparison popular descriptors transform and speeded‐up robust features, authors also choose detected by difference Gaussian fast Hessain...

To improve the accuracy of human pose estimation, a novel method based on deep high-resolution network (HRNet) and equipped with double attention residual blocks is proposed. Firstly, channel spatial modules are added to block feature extraction, resulting in paying more target area which needs be extracted important information suppressed unimportant information. Moreover, this paper proposes module, Parallel Residual Attention Block (PRAB), parallels 3 × group convolution ResNeXt layer...

IEEE 802.16e is the latest broadband wireless access standarddesigned to support mobility. In mobile networks, how controlenergy consumption one of most important issues for thebattery-powered stations. The standard proposes an energysaving mechanism that named 'sleep mode' conserving powerof According operation ofthe sleep mode downlink traffic in type I power savingclass, a discrete-time Geom/G/1 queueing model with close-downtime and multiple vacations built. By employing embeddedMarkov...

IEEE 802.16e is the latest broadband wireless access standard designed to support mobility. In mobile networks, how control energy consumption one of most important issues for battery-powered stations. The proposes an saving mechanism named "sleep mode" conserving power According operating sleep mode downlink traffic in type I class, a discrete-time Geom/G/1 queueing model with close-down time and multiple vacations built this paper. By employing embedded Markov chain method, average queue...

Wireless regional area network (WRAN) adopts centralized architecture and is currently one of the most typical cognitive radio networks. In order to reduce energy consumption communication networks with constraint spectrum resource utilization, a working sleep mechanism introduced into base station (BS), novel saving strategy dual rate transmission proposed. Combining multiple-vacation queue priority queue, using quasi-birth-death process matrix-geometric solution method, we assess average...

This study investigates the problem of constructing binary descriptor and develops a novel called simple tri‐bit (STBD) based on sampling pattern (SSP) tri‐value binarisation strategy (TBS). First, an SSP is proposed, in which sample points are divided into two groups according to distance from centre smoothed by different circular filters. Then, make adaptive matched images, selection directly employs detected keypoints as training data introduced select 256 point pairs with low correlation...

Crowd counting is a task that aims to estimate the number of people in an image. Recent crowd methods make significant progress by employing convolutional neural networks regress density maps. One most challenging problems this drastic scale variation region interest images. In paper, Feature Fusion Attention Network (FFANet) proposed for counting. Firstly, VGG16 network adapted as backbone FFANet extract features Then, extracted are fused subsequent two stages. Specifically, information...