A5101826440- 3D Surveying and Cultural Heritage
- Image and Object Detection Techniques
- Indoor and Outdoor Localization Technologies
- Speech and Audio Processing
- Image Processing and 3D Reconstruction
- 3D Shape Modeling and Analysis
- Advanced Vision and Imaging
- Advanced Neural Network Applications
- Medical Image Segmentation Techniques
- Advanced Image Processing Techniques
- Millimeter-Wave Propagation and Modeling
- Computer Graphics and Visualization Techniques
- Wireless Networks and Protocols
- Image Processing Techniques and Applications
- Industrial Vision Systems and Defect Detection
- Semiconductor Lasers and Optical Devices
- Photonic and Optical Devices
- Optical measurement and interference techniques
- Underwater Vehicles and Communication Systems
- Infrastructure Maintenance and Monitoring
- Remote Sensing and LiDAR Applications
- Advanced Fiber Optic Sensors
- Neural Networks and Reservoir Computing
- Mechanical and Optical Resonators
- Quantum Information and Cryptography
University of Missouri
2016-2020
Ho Chi Minh City University of Science
2011-2013
Vietnam National University Ho Chi Minh City
2011
Nanyang Technological University
2009
This paper reviews the first challenge on single image super-resolution (restoration of rich details in an low resolution image) with focus proposed solutions and results. A new DIVerse 2K dataset (DIV2K) was employed. The had 6 competitions divided into 2 tracks 3 magnification factors each. Track 1 employed standard bicubic downscaling setup, while unknown operators (blur kernel decimation) but learnable through high res train images. Each competition ∽100 registered participants 20 teams...
Volumetric grid is widely used for 3D deep learning due to its regularity. However the use of relatively lower order local approximation functions such as piece-wise constant function (occupancy grid) or linear (distance field) approximate shape means that it needs a very high-resolution represent finer geometry details, which could be memory and computationally inefficient. In this work, we propose PointGrid, convolutional network incorporates number points within each cell thus allowing...
In this paper, we introduce REDN: A Recursive Encoder-Decoder Network with Skip-Connections for edge detection in natural images. The proposed network is a novel integration of Neural an architecture. recursive enables iterative refinement the edges using single model. Adding skip-connections between encoder and decoder helps gradients reach all layers more easily allows information related to finer details early stage be fully utilized decoder. Based on our extensive experiments popular...
Circle detection from digital images is a necessary operation in many robotics and computer vision tasks to facilitate shape object recognition. We propose analyze novel method, based on line segment circle completeness verification, detect circles images. The key idea use segments instead of raw edge pixels get the candidates followed by verification step measure circle's completeness. Experimental results several synthesized hand-sketched as well natural with various complication favor...
This paper introduces a method which works together with the fingerprint localization algorithm in order to make it more applicable many different mobile devices. The original performs effectively only on sample device used build map training phase. Thus, full is needed for each new and this process time-consuming (few hours or more). We propose an approach that can avoid requires quick calibration step (a few minutes). When tested algorithm, our achieves accuracy of 69% experiment was...
Rock detachment events along roadways pose public safety concerns but can be predicted and safely handled using geological measurements of discontinuities. With modern sensing technology, these taken on 3-D point clouds 2-D optical images that provide a high level structural accuracy visual detail. Doing so allows engineers to obtain the needed data with relative ease while eliminating biases hazards inherent in taking manual measurements. This letter presents an approach for fusing natural...
LIDAR range scans can be used to quickly create accurate 3D models for virtual reality and as a basis visualize sets of photographs, videos, objects in cohesive environment. The number existing programs that use data input has motivated our group develop methods fusing images with augmenting the both dynamic present videos models. Bringing together many sources possible increases users' abilities related information one, intuitive venue. We demonstrate how register variety 2D imagery scan...
This paper proposes a scalable method which allows deploying the fingerprint Wi-Fi localization algorithm for different devices. The original performs accurately only if device used in testing phase is same as training phase. When phase, time-consuming re-training step (in order of hours or days) required to achieve equivalent degree accuracy. Our proposed approach replaces with short period calibration few minutes), can be done transparently user. To validate our approach, we collected data...
As GPS signals may not be available indoor and can blocked easily by obstacle, several localization techniques based on Wi-Fi have been proposed. Most of them require a pre-training step for known locations which is time-consuming practical wide-scale deployment. This paper proposes an alternative approach the SSID nearby access points (APs) at specific location. The idea this work observation that many AP's names contain information related to Thus, observable APs' are mined via search...
Recent research suggests that quantum radar offers several potential advantages over classical sensing technologies. At present, the primary practical challenge is fast and efficient generation of entangled microwave photons. To mitigate this limitation we propose briefly examine a distributed architecture to synthetically increase number effectively-distinguishable modes.
In this paper, we introduce RED-NET: A Recursive Encoder-Decoder Network with Skip-Connections for edge detection in natural images. The proposed network is a novel integration of Neural an architecture. recursive enables us to increase the depth without increasing number parameters. Adding skip-connections between encoder and decoder helps gradients reach all layers more easily allows information related finer details early stage be fully utilized decoder. Based on our extensive experiments...
This paper proposes an applicable and scalable approach which allows deploying the fingerprint Wi-Fi localization algorithm for different mobile devices. The original performs accurately when device used in deployment phase is same as training phase. However, a phase, time-consuming re-training step (in order of hours or days) required to achieve equivalent degree accuracy. Our proposed replaces this by short period calibration few minutes), can be done transparently user. To validate our...
We demonstrate that a prism coupling characterization technique can lead to simpler Waveguide Bragg Grating (WBG) designs and easier fabrication by avoiding channel transmission. The experimental results have shown the incident laser beam's collimation was maintained after planar waveguide with an overall efficiency of 10%.