RGB-D salient object detection (SOD) is usually formulated as a problem of classification or regression over two modalities, i.e., RGB and depth. Hence, effective feature modeling multi-modal fusion both play vital role in SOD. In this paper, we propose depth-sensitive scheme using the depth-wise geometric prior objects. principle, carried out attention module, which leads to enhancement well background distraction reduction by capturing depth geometry prior. More-over, perform fusion,...

The development of microarray-based high-throughput gene profiling has led to the hope that this technology could provide an efficient and accurate means diagnosing classifying tumors, as well predicting prognoses effective treatments. However, large amount data generated by microarrays requires reduction discriminant features into reliable sets tumor biomarkers for such multiclass discrimination. availability biomarkers, especially serum should have a major impact on our understanding...

Designing a lightweight semantic segmentation network often requires researchers to find trade-off between performance and speed, which is always empirical due the limited interpretability of neural networks. In order release from these tedious mechanical trials, we propose Graph-guided Architecture Search (GAS) pipeline automatically search real-time Unlike previous works that use simplified space stack repeatable cell form network, introduce novel mechanism with new where model can be...

The detection of arbitrary-oriented and multi-scale objects in satellite optical imagery is an important task remote sensing computer vision. Despite significant research efforts, such remains largely unsolved due to the diversity patterns orientation, scale, aspect ratio, visual appearance; dense distribution objects; extreme imbalances categories. In this paper, we propose adaptive dynamic refined single-stage transformer detector address aforementioned challenges, aiming achieve high...

Wood surface defect detection is a critical step in wood processing and manufacturing. To address the performance degradation caused by small targets multi-scale features detection, novel deep learning model proposed this study, FDD-YOLO, specifically designed for task. In feature extraction stage, C2f module funnel attention (FA) mechanisms are integrated into design of C2f-FA to enhance model’s ability extract defects various sizes. Additionally, Dual Spatial Pyramid Pooling-Fast (DSPPF)...

Tree species surveys are crucial in forest resource management and can provide references for protection policymakers. Traditional tree the field labor-intensive time-consuming. In contrast, airborne LiDAR technology is highly capable of penetrating vegetation; it be used to quickly obtain three-dimensional information regarding vegetation over large areas with a high level precision, widely forestry. At this stage, most studies related individual classification focus on traditional machine...

Traditional nursery seedling detection often uses manual sampling counting and height measurement with rulers. This is not only inefficient inaccurate, but it requires many human resources for nurseries that need to monitor the growth of saplings, making difficult meet fast efficient management requirements modern forestry. To solve this problem, paper proposes a real-time framework based on an improved YoloV4 network binocular camera, which can provide measurements number saplings in...

<title>Abstract</title> Using deep learning methods is a promising approach to improving bark removal efficiency and enhancing the quality of wood products. However, lack publicly available datasets for plate segmentation in processing poses challenges researchers this field. To address issue, benchmark named WPS-dataset proposed study, which consists 4863 images. We designed an image acquisition device assembled it on equipment capture images real industrial settings. evaluated using six...

The 3D reconstruction of point cloud trees and the acquisition stand factors are key to supporting forestry regulation urban planning. However, two usually independent modules in existing studies. In this work, we extended AdTree method for modeling by adding a quantitative analysis capability acquire factors. We used unmanned aircraft LiDAR (ALS) data as raw study. After denoising segmenting single trees, obtained single-tree samples needed study produced our own sample dataset. scanned...

We propose a robust approach to detecting and tracking moving objects for naval unmanned aircraft system (UAS) landing on an carrier. The frame difference algorithm follows simple principle achieve real-time tracking, whereas Faster Region-Convolutional Neural Network (R-CNN) performs highly precise detection characteristics. thus combine R-CNN with the method, which is demonstrated exhibit performance. In our UAS experiments, two cameras placed both sides of runway are used capture UAS....

Currently, most methods in clinical psychology research primarily rely on questionnaires and interviews with examiners, which could not provide real-life subject behavioral data monitoring collecting services. This paper presents a smartphone-based mobile Ambulatory Assessment System, called mAAS, for research, especially alcohol craving studies, to improve current real-time collecting. system consists of wearable sensor, Equivital EQ2 measuring physiological data, an Android smartphone, web...

Mobile ambulatory assessment systems are being actively developed for various psychological studies. However, to the best of our knowledge, there very few used detection alcohol usage. This paper presents a new automatic data analysis pipeline, called ADA (Automatic Detection Alcohol usage), assess use or cravings and emotional dysregulation associated with underlying disorders. works closely mAAS, smartphone-based mobile system psychology research, currently by approximately 30 subjects in...

Template matching is an important and challenging task in remote sensing computer vision. Existing template methods often fail the presence of complex nonrigid deformation, occlusion, background clutter. In this letter, inspired by Siamese trackers, we propose end-to-end method that based on network. Different from traditional methods, our treats as a classification-regression task. It more robust to clutter, deformation. Moreover, introduce channel-attention mechanism cross correlation...

In this paper, a systematic description of the artificial intelligence (AI)-based channel estimation track 2nd Wireless Communication AI Competition (WAIC) is provided, which hosted by IMT-2020(5G) Promotion Group 5G+AI Work Group. Firstly, system model demodulation reference signal (DMRS) based problem and its corresponding dataset are introduced. Then potential approaches for enhancing performance discussed from viewpoints data analysis, pre-processing, key components backbone network...

Arbitrary-oriented object detection is a relatively emerging but challenging task. Although remarkable progress has been made, there still remain many unsolved issues due to the large diversity of patterns in orientation, scale, aspect ratio, and visual appearance objects aerial images. Most existing methods adopt coarse-grained fixed label assignment strategy suffer from inconsistency between classification score localization accuracy. First, align metric sample selection regression loss...

Based on the generalized Huygens-Fresnel diffraction integral (Collins' formula), propagation equation of Hermite-Gauss beams through a complex optical system with limiting aperture is derived. The elements may be all those characterized by an ABCD ray-transfer matrix, as well any kind apertures represented transmittance functions. To obtain analytical expression, we expand function into finite sum Gaussian Thus expressed superposition series Gaussian-shaped apertures. advantage this...

Aiming at the characteristics of large changes in object scale and complex background urban aerial image, we propose an advanced YOLOv3 detection algorithm to solve it. First, analyze data through a clustering calculate optimal size prior anchors. Then, relatively lightweight easily extensible backbone network-deep residual network is used for feature extraction. Meanwhile, order obtain better receptive field further reduce information loss process convolution, add deformable convolution 8...