A5101529572- Robotics and Sensor-Based Localization
- Organic Electronics and Photovoltaics
- Advanced Vision and Imaging
- Conducting polymers and applications
- Advanced Photocatalysis Techniques
- 3D Surveying and Cultural Heritage
- 3D Shape Modeling and Analysis
- Perovskite Materials and Applications
- Metal-Organic Frameworks: Synthesis and Applications
- Robot Manipulation and Learning
- Covalent Organic Framework Applications
- Hydrology and Watershed Management Studies
- Advanced Image and Video Retrieval Techniques
- Organic Light-Emitting Diodes Research
- Remote Sensing and LiDAR Applications
- Image and Object Detection Techniques
- Quantum Dots Synthesis And Properties
- Electrochemical sensors and biosensors
- Flood Risk Assessment and Management
- Silicon Nanostructures and Photoluminescence
- Advanced Semiconductor Detectors and Materials
- Advanced Neural Network Applications
- Smart Grid Energy Management
- Image Processing and 3D Reconstruction
- Remote-Sensing Image Classification
Huazhong University of Science and Technology
2022-2025
Nanjing University
2024
Huawei Technologies (China)
2022-2024
Changzhou University
2021-2024
Hong Kong Polytechnic University
2022
Northwest Normal University
2022
Huawei Technologies (France)
2021
Stanford University
2019
New York University
2018
Tel Aviv University
2018
The goal of this paper is to estimate the 6D pose and dimensions unseen object instances in an RGB-D image. Contrary "instance-level'' estimation tasks, our problem assumes that no exact CAD models are available during either training or testing time. To handle different a given category, we introduce Normalized Object Coordinate Space (NOCS)-a shared canonical representation for all possible within category. Our region-based neural network then trained directly infer correspondence from...
The large non-radiative recombination is the main factor that limits state-of-the-art organic solar cells (OSCs). In this work, two novel structurally similar oligomers (named 5BDTBDD and 5BDDBDT) with D-A-D-A-D A-D-A-D-A configuration are synthesized for high-performance ternary OSCs low energy loss. As third components, these PM6 analogue effectively suppress in OSCs. Although highest occupied molecular orbital (HOMO) levels of 5BDDBDT higher than PM6, enabled ultra-high...
We present PrimitiveNet, a novel approach for high-resolution primitive instance segmentation from point clouds on large scale. Our key idea is to transform the global problem into easier local tasks. train embedding network predict explicit geometry features and implicit latent each point. The jointly trained with an adversarial as discriminator decide whether points are same in neighborhoods. Such supervision encourages learned describe surface properties robustly distinguish different...
We present HoliCity, a city-scale 3D dataset with rich structural information. Currently, this has 6,300 real-world panoramas of resolution $13312 \times 6656$ that are accurately aligned the CAD model downtown London an area more than 20 km$^2$, in which median reprojection error alignment average image is less half degree. This aims to be all-in-one data platform for research learning abstracted high-level holistic structures can derived from city models, e.g., corners, lines, wireframes,...
This technical report presents the 1st winning model for UG2+, a task in CVPR 2024 UAV Tracking and Pose-Estimation Challenge. challenge faces difficulties drone detection, UAV-type classification 2D/3D trajectory estimation extreme weather conditions with multi-modal sensor information, including stereo vision, various Lidars, Radars, audio arrays. Leveraging this we propose classification, 3D tracking method accurate tracking. A novel pipeline which incorporates sequence fusion, region of...
Reservoir operation plays an important role in reservoir management. In operation, water balance calculation is a very step. At present, one of the main challenges that inflow cannot be calculated accurately due to jacking reservoir, which produced by downstream after original course river has changed. Another reason capacity influence dynamic storage capacity. order overcome these problems, this report shows land zone front dam can used calculate capacity, since it serve as boundary for...
We present a novel vectorized indoor modeling approach that converts point clouds into building information models (BIM) with concise and semantically segmented polygonal meshes. Existing methods detect planar shapes connect them to complete the scene. Some focus on floor plan reconstruction as simplified problem better analyze connectivity between planes of floors walls. However, analysis is still challenging ill-posed incomplete input. propose ArrangementNet estimate scene arrangements...
We present MVLayoutNet, a network for holistic 3D reconstruction from multi-view panoramas. Our core contribution is to seamlessly combine learned monocular layout estimation and stereo (MVS) accurate in both image space. jointly train module produce an initial novel MVS obtain geometry. Unlike standard MVSNet, our takes newly-proposed cost volume, which aggregates costs at the same depth layer into corresponding elements. additionally provide attention-based scheme that guides focus on...
Abstract In the fabrication of hafnium oxide (HfO 2 )-based metal–insulator–semiconductor (MIS) devices, a sacrificial amorphous silicon (a-Si) film was used as source for facilitating formation silicate (Hf-silicate; HfSiO) between HfO and crystallized Si (c-Si). HfSiO can assist in changing phase into tetragonal achieve high dielectric constant. The combination named “Hybrid ”. When this Hybrid insulator applied to MIS it form good insulator/semiconductor interface with c-Si. cannot only...
External quantum efficiency (EQE), color rendering index (CRI), correlated temperature (CCT), and spectral stability are key challenges for constructing high-performance white organic light-emitting diodes (WOLEDs) with a simple device structure. Herein, in order to resolve these issues, we primarily employed thermally activated delayed fluorescence (TADF) emitter of CzAcSF as host dinuclear platinum (III) complex (Pt-1) guest construct solution-processed single-emitting layer WOLEDs at low...
Abstract We propose a unified 3D flow framework for joint learning of shape embedding and deformation different categories. Our goal is to recover shapes from imperfect point clouds by fitting the best template in repository after deformation. Accordingly, we learn retrieval flow-based network robust note that can be quite Therefore, introduce novel multi-hub module multiple modes incorporate such variation, providing which handle wide range objects The designed retrieve best-fit as nearest...