We present Vid2Avatar, a method to learn human avatars from monocular in-the-wild videos. Reconstructing humans that move naturally videos is difficult. Solving it requires accurately separating arbitrary backgrounds. Moreover, reconstructing detailed 3D surface short video sequences, making even more challenging. Despite these challenges, our does not require any groundtruth supervision or priors extracted large datasets of clothed scans, nor do we rely on external segmentation modules....

We present a novel method to learn Personalized Implicit Neural Avatars (PINA) from short RGB-D sequence. This allows non-expert users create detailed and personal-ized virtual copy of themselves, which can be animated with realistic clothing deformations. PINA does not require complete scans, nor it prior learned large datasets clothed humans. Learning avatar in this setting is challenging, since only few depth observations are available, noisy incomplete (i.e. partial visibility the body...

According to literature, myopia has become the second most common eye disease in China, and incidence of is increasing year by year, showing a trend younger age. Previous researches have shown that occurrence mainly determined poor habits, including reading writing posture, length, so on, parents’ heredity. In order better prevent adolescents, this paper studies influence related factors on adolescents based machine learning method. A feature selection method both univariate correlation...

Abstract The type recognition of a weld defect in an ultrasonic Time Flight Diffraction (TOFD) image is lack efficiency, stability, and reliability that due to the limitation experience professional knowledge inspector. In this study, characteristics TOFD D-scan images were analyzed, Faster Region-based Convolutional Neural Networks (Faster R-CNN) was adopted for autorecognition defect’s type. course training process, proposed box configuration optimized improve imaging samples expanded...

Most imaging devices lose image information during the acquisition process due to their low dynamic range (LDR). Existing high (HDR) techniques have a trade-off with time or spatial resolution, resulting in potential motion blur misalignment. Current HDR methods are based on fusion of multi-frame LDR images and can suffer from blurring fine details, aliasing, boundary effects. This study developed dual-channel camera (DCC) achieve imaging, which eliminate registration problems. Considering...

Generally, the dynamic range of night vision scenes is large. Owing to limited traditional low light imaging technology, captured images are always partially overexposed or underexposed. Multi-exposure fusion most effective method overcoming limitation sensor, and multi-frame (LDR) image a key consideration. However, existing methods have problems such as detail blurring aliasing. This paper proposes an multi-scale decomposition based on gradient domain guided filter (GDGF), which can better...

The dynamic range of night vision scenes is typically very large. Owing to the limited traditional low-light-level imaging technology, captured images are always partially overexposed or underexposed. Multi-exposure fusion most effective method for overcoming limitations sensors. Recently, deep learning has achieved tremendous progress in many fields. However, only a few breakthroughs have been reported on high-dynamic image with method. Additionally, problems conjunctions commonly used...

Weakly supervised object detection (WSOD) is an increasingly important task in remote sensing images. However, mainstream WSOD methods often rely on low-quality proposals due to the complex backgrounds of Moreover, applying strong data augmentations directly can introduce significant noise, which hinder training procedures that only image-level ground truth labels. To address these issues, we propose a dynamic informative proposal-based iterative procedure. Specifically, implement...

At the present time, there are many image contrast enhancement methods where main considerations detail enhancement, noise suppression, and high suppression. Traditional ignore characteristics of display or merely consider as a whole. However, due to limited dynamic range most devices on market, difference between two adjacent grayscales is often below just noticeable human visual systems, which causes details be invisible display. To solve this problem, we preprocessing method for...

An efficient algorithm for automatic road extraction is presented based on fuzzy mask in this paper. The concepts are employed order to perform pixels extraction. aim of the proposed approach attribute a membership value each pixel. Firstly, set sets, which represent basic structures, defined. structure, defined as logical concatenation these elementary predefined 2D used enhance values. input data SPOT panchromatic band remote sensing required, and it applied system. Obtained experimental...

With the pursuit of a balance between industrial development and environmental protection, importance natural elements in application materials is becoming increasingly prominent. The article mainly focuses on definition classification materials, as well their mechanism action improving material properties. analyzes unique contributions plant-based mineral bio-based polymers enhancing Meanwhile, also points out cost, technological, regulatory challenges faced by applications, innovative...

Abstract High-speed hydraulic dynamometer is widely used for turbine component experiment in aircraft engine area. It usually works through measuring main performance parameters of verifying design method by absorbing huge shaft work which passed from turbine. However, its water flow can generate strong turbulence phenomena between impellers dynamometer’s case. In addition, high temperature may cavitation phenomenon and cause high-frequency pressure pulsation. Both lead to degradation....

Abstract Accurately predicting the thrust of aero-engine is important for engine health management and performance control. Thrust prediction neural network models have been widely applied in recent years. This paper proposes a parameter augmentation pretraining method designs an model with physical significance. It known that ground test system measures more parameters compared to onboard system. Due limitations using models, typically only measured by both systems are used as training...

In this paper, an end-to-end learning framework—RGB enhanced point cloud fusion network (REPF-Net) is designed based on the complementary and of multi-sensor (camera lidar) data, aiming to realize accurate perception robot in a random complex environment. other words, three-dimensional position geometric shape environmental targets can be obtained excellently accuracy target recognition improved. Meanwhile, three key problems 3D tasks were solved including, enhancing with RGB feature points,...

Chronic diseases represented by circulatory have gradually become the main types of affecting health our population. Establishing a system disease prediction model to predict occurrence and controlling them is great significance This article based on prospective population cohort data chronic in China, existing medical studies, Kaplan–Meier method was used for feature selection, traditional analysis Cox proportional hazards introduced. Support vector machine research methods learning...

With the continuous improvement of number and capability micro-nano satellites, on-board intelligent data processing becomes a necessary configuration. The constellation with hundreds satellites has ability to high-frequency detection ship targets realizes awareness global ocean, which is great significance in maritime rescue, waterway management combating illegal fishing. In this paper, fast method for panchromatic image proposed. Firstly, GPU (graphics unit) commercial devices used form...

Abstract Integrated navigation systems made up of a strap-down inertial system (SINS) and global positioning (GPS) are increasingly being used to improve the position, speed, attitude information unmanned surface vessels (USV). However, GPS outage could occur due dependence performance on external environment number available satellites. This study uses an innovative combination Dempster–Shafer (DS) theory broad learning (BL) method design SINS/GPS integrated system. First, velocity position...