It is a challenging task to recognize smoke from images due large variance of color, texture, and shapes. There are detection methods that have been proposed, but most them based on hand-crafted features. To improve the performance detection, we propose novel deep normalization convolutional neural network (DNCNN) with 14 layers implement automatic feature extraction classification. In DNCNN, traditional replaced accelerate training process boost detection. reduce overfitting caused by...

Abstract Background Protein subcellular localization plays a crucial role in understanding cell function. Proteins need to be the right place at time, and combine with corresponding molecules fulfill their functions. Furthermore, prediction of protein location not only should guiding drug design development due potential molecular targets but also an essential genome annotation. Taking current status image-based as example, there are three common drawbacks, i.e., obsolete datasets without...

Abstract In the field of small object detection, Yolov4-Tiny is inadequate in feature extraction and does not make best multi-scale features. this paper, an improved BiFPN framework proposed based on to increase detection precision. Moreover, taken as backbone network introduce spatial pyramid pooling (SPP) connect merge regions. Finally, our method can achieve 79.53% map Pascal VOC dataset, which 2.12% higher than original model.

Face detection is one of the most basic tasks in many various face applications, which gradually becoming acceptable biometric recognition method. However, tinyface a complex and challenging pattern problem that encounters difficulties application process. This paper uses popular deep learning algorithm to complete tiny-face combine Yolov3-tiny model with Dropblock strategy. regularization method used convolutional layer, activation units are spatially interrelated. Neurons contiguous region...

With the rapid development of Internet, especially extensive application deep learning, change detection has been successfully applied in many fields, but pursuit greater feature information content led to an increase memory and computing power requirements. To solve this problem, a Lightweight Siamese attention Residual Network (LSRNet) is proposed paper reduce computational requirements, embed Fast Small Attention (SFSAttention) filter out with less relevance, then use fusion channel...

Abstract In the past ten years, deep learning has achieved remarkable results in area of natural image segmentation, and gradually turned to field medical segmentation. The precise segmentation spine images can be used for early screening spondylopathy, which is convenient detection treatment patients. Aiming at by U-Net network, structure will lead large model calculation, network overfitting, size, noise information other issues. This paper introduces a new method based on spatial pyramid...

Cancer is one of the serious diseases, recent studies reported that tumor homing peptides (THPs) play a key role in treatment cancer. Due to experimental methods are time-consuming and expensive, it urgent develop automatic computational approaches identify THPs. Hence, this study, we proposed novel machine learning distinguish THPs from non-THPs, which peptide sequences firstly encoded by pseudo residue pairwise energy content matrix (PseRECM) physicochemical property (PsePC). Moreover,...

Research on video analysis and processing of fire smoke detection has gradually become a popular topic in computer vision. It is also challenging task to detect videos due the non-rigid characteristics large variance color, texture, shape, density lighting, causing most existing video-based algorithms with high false rate. In this paper, we combine Gaussian Mixture Model (GMM) HSV color model deep convolution for detecting smoke, which aim at filtering out no-smoke blocks further reduce rate...

Object detection task, as a prevailing direction of computer vision, involves many challenges. One its most general problems is overfitting. Random Erasing state-of-art Data Augmentation method for avoiding However, it aims at classification task. When used to train object models, sometimes discards the objects, then bounding boxes correspond some noise regions. To solve this shortcoming Erasing, paper proposes Range-aware data augment method. In training stage, randomly occludes part...

Recently, accurate target tracking is widely used in the field of Unmanned Aerial Vehicles (UAV). In this paper, we focus on application detecting and following a walking pedestrian real time from moving platform with many interferences. We present scheme that uses CNN model (YOLO-V2) to detect matches postprocessing feature queue Locality constrained Linear Coding algorithm. After ground station receives analyses video stream parrot sends back commands control motion UAV. At beginning...

Recently, visual tracking has been widely concerned in computer vision applications such as behavior analysis, autonomous driving with advanced trackers. Some trackers based on Discriminative Correlation Filters (DCF) have acquired great performance benchmark, but it is not satisfied the real-time tracking; Later, some people combined Deep Neural Networks (DNNs) and DCF to form trackers, which significantly improved accuracy speed. Unfortunately, cannot be applied long-term many public...

Recent studies reported that N7-methylguanosine (m7G) plays a vital role in gene expression regulation. As consequence, determining the distribution of m7G is crucial step towards further understanding its biological functions. Although experimental approaches are capable accurately locating sites, they labor-intensive, costly, and time-consuming. Therefore, it necessary to develop more effective robust computational methods replace, or at least complement current methods. In this study, we...

Smoke accurate detection, for the real-time fire detection and early warning has an important role.In order to overcome problem that smoke is low when burned, video method based on local binary mode proposed under condition of disturbing wind speed other factors.In this method, motion region extracted by background subtraction each piece area processed obtain information.Then, texture feature block using model.Finally, features are used feature.To achieve image extraction.Finally, support...

The primary issue in the study of oracle is how to recognize and understand characters. Therefore, recognition an significant research field. In this paper, we firstly exploy different convolutional neural network (CNN) architectures, which are aim at implementing automatic feature extraction classification improve performance recognition. We also show that Gabor method, demonstrated effectively precision classification. To reduce overfitting caused by insufficient training samples, generate...

Synthetic Aperture Radar (SAR) imaging plays a vital role in maritime surveillance, yet the detection of small vessels poses significant challenge when employing conventional Constant False Alarm Rate (CFAR) techniques, primarily due to limitations resolution and presence clutter. Deep learning (DL) offers promising alternative, it still struggles with identifying targets complex SAR backgrounds because feature ambiguity noise. To address these challenges, our team has developed AFSC...

This paper deals with the video surveillance problem in static camera. The aim is to design a moving object detection system based on advanced CPU DM6437. After capturing pictures according timing, background model established without objects and information real-time updated due changes of environment. that, subtraction used so as figure out objects. morphological methodology area marking are utilized for eliminating noise. Experiment results state that can identify accurately under error...