A5022442335- Advanced Vision and Imaging
- Robotics and Sensor-Based Localization
- Video Surveillance and Tracking Methods
- Remote Sensing and LiDAR Applications
- 3D Surveying and Cultural Heritage
- Computational Drug Discovery Methods
- Advanced Image and Video Retrieval Techniques
- Machine Learning in Materials Science
- Cardiac electrophysiology and arrhythmias
- Anesthesia and Neurotoxicity Research
- Molecular Biology Techniques and Applications
- Advanced Nanomaterials in Catalysis
- Teleoperation and Haptic Systems
- Infrared Target Detection Methodologies
- Advanced Neural Network Applications
- Nanoparticles: synthesis and applications
- Machine Learning in Bioinformatics
- Machine Learning and Data Classification
- Carcinogens and Genotoxicity Assessment
- Image Retrieval and Classification Techniques
- Automated Road and Building Extraction
- Evolutionary Algorithms and Applications
University of Tsukuba
2024
Tianjin University
2020
University of Alberta
2017-2018
Aims: The study aims to find a way reduce the dimensionality of dataset. Background: Dimensionality reduction is key issue machine learning process. It does not only improve prediction performance but also could recommend intrinsic features and help explore biological expression “black box”. Objective: A variety feature selection algorithms are used select data achieve reduction. Methods: First, MRMD2.0 integrated 7 different popular ranking with PageRank strategy. Second, optimized was...
This paper presents a novel boundary based semiautomatic tool, ByLabel, for accurate image annotation. Given an image, ByLabel first detects its edge features and computes high quality fragments. Current labeling tools require the human to accurately click on numerous points. simplifies this just selecting among fragment proposals that automatically generates. To evaluate performance of By-Label, 10 volunteers, with no experiences annotation, labeled both synthetic real images. Compared...
This letter presents a novel approach for extracting accurate outlines of individual buildings from very high-resolution (0.1-0.4 m) optical images. Building are defined as polygons here. Our operates on set straight line segments that detected by detector. It groups subset and connects them to form closed polygon. Particularly, new grouping cost is first. Second, weighted undirected graph G(V,E) constructed based the endpoints those extracted segments. The building outline extraction then...
Although semi-dense Simultaneous Localization and Mapping (SLAM) has been becoming more popular over the last few years, there is a lack of efficient methods for representing processing their large scale point clouds. In this paper, we propose using 3D line segments to simplify clouds generated by SLAM. Specifically, present novel incremental approach segment extraction. This reduces fitting problem into two 2D problems takes advantage both images depth maps. our method, are fitted...
During drug development, ensuring that molecules do not block the hERG (human Ether-à-go-go-Related Gene) channel is critical. If this blocked, it can cause many cardiovascular-related diseases. However, traditional experimental detection methods are expensive and time-consuming. In work, we proposed a novel deep learning framework CLOP-hERG combines contrastive with RoBERTa pre-trained model to predict whether will channel. We employed data augmentation techniques on molecular structures...
This paper presents a novel real-time method for tracking salient closed boundaries from video image sequences. operates on set of straight line segments that are produced by detection. The scheme is coherently integrated into perceptual grouping framework in which the visual problem tackled identifying subset these and connecting them sequentially to form boundary with largest saliency certain similarity previous one. Specifically, we define new criterion combines cost an area constraint....
This paper presents a novel real-time method for tracking salient closed boundaries from video image sequences. operates on set of straight line segments that are produced by detection. The scheme is coherently integrated into perceptual grouping framework in which the visual problem tackled identifying subset these and connecting them sequentially to form boundary with largest saliency certain similarity previous one. Specifically, we define new criterion combines cost an area constraint....
During drug development, ensuring that molecules do not block the hERG (human Ether-à-go-go-Related Gene) channel is critical. If this blocked, it can cause many cardiovascular-related diseases. However, traditional experimental detection methods are expensive and time-consuming. In work, we proposed a novel deep learning framework CLOP-hERG, combines contrastive with RoBERTa pre-trained model to predict whether will channel. We employed data augmentation techniques on molecular structures...
Although semi-dense Simultaneous Localization and Mapping (SLAM) has been becoming more popular over the last few years, there is a lack of efficient methods for representing processing their large scale point clouds. In this paper, we propose using 3D line segments to simplify clouds generated by SLAM. Specifically, present novel incremental approach segment extraction. This reduces fitting problem into two 2D problems takes advantage both images depth maps. our method, are fitted...
In this paper, we propose a novel real-time method for tracking planar edge templates. This tracks an template by estimating its homography transformations with respect to the sampled pixels detected from incoming frames. Particularly, define cost function based on new feature map of to-be-tracked and optimize it Lucas-Kanade-like algorithm. The is defined as fourth root distance transform. Our operates just edges so that good at those low textured targets, such hollow targets (mug rim),...
We present a coarse-to-fine approach based semi-autonomous teleoperation system using vision guidance. The is optimized for long range tasks under time-delay network conditions and does not require prior knowledge of the remote scene. Our initializes with self exploration behavior that senses surroundings through freely mounted eye-in-hand web cam. stage estimates hand-eye calibration provides telepresence interface via real-time 3D geometric reconstruction. human operator able to specify...