Tuấn Anh Nguyễn - Academia.edu (original) (raw)
Uploads
Papers by Tuấn Anh Nguyễn
Journal of Ambient Intelligence and Humanized Computing, 2013
One of the key inputs for building energy and comfort management systems is detailed and precise ... more One of the key inputs for building energy and comfort management systems is detailed and precise office activity information, i.e., what is happening in each of the monitored areas (e.g., working place, meeting room, coffee corner, etc.). Such information can be used to control appliances for saving energy while still satisfying user comfort needs accordingly. We propose an activity recognition solution that effectively handles multiple-user, multiple-area situations, rapidly recognizing office activities as inputs for building energy and comfort management systems. The proposed solution is based on an ontological approach, using lowcost, binary, and wireless sensors. Through initial experiments, our prototype is able to recognize seven typical office activities (working at a desk with or without a PC, having a meeting, giving a presentation, having a coffee break, and presence/absence) at three office activity areas (working room, meeting room, and coffee corner) for two persons with average accuracy of more than 92%.
Energy and Buildings, 2013
Take-down policy If you believe that this document breaches copyright please contact us providing... more Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
2016 23rd International Conference on Pattern Recognition (ICPR), 2016
The encoding method is an important factor for an action recognition pipeline. One of the key poi... more The encoding method is an important factor for an action recognition pipeline. One of the key points for the encoding method is the assignment step. A very widely used super-vector encoding method is the vector of locally aggregated descriptors (VLAD), with very competitive results in many tasks. However, it considers only hard assignment and the criteria for the assignment is performed only from the features side, by looking for which visual word the features are voting. In this work we propose to encode deep features for videos using a double assignment VLAD (DA-VLAD). In addition to the traditional assignment for VLAD we perform a second assignment by taking into account the perspective from the codebook side: which are the nearest features to a visual word and not only which is the nearest centroid for the features as the standard assignment. Another important factor for the performance of an action recognition system is the feature extraction step. Recently, deep features obtained state-of-the-art results in many tasks, being also adopted for action recognition with competitive results over handcrafted features. This work includes a pipeline to extract local deep features for videos using any available network as a black box and we show competitive results including the case when the network was trained for another task or another dataset. Our DA-VLAD encoding method outperforms the traditional VLAD and we obtain state-of-the-art results on UCF50 dataset and competitive results on UCF101 dataset.
2016 14th International Workshop on Content-Based Multimedia Indexing (CBMI), 2016
Besides appearance information, the video contains temporal evolution, which represents an import... more Besides appearance information, the video contains temporal evolution, which represents an important and useful source of information about its content. Many video representation approaches are based on the motion information within the video. The common approach to extract the motion information is to compute the optical flow from the vertical and the horizontal temporal evolution of two consecutive frames. However, the computation of optical flow is very demanding in terms of computational cost, in many cases being the most significant processing step within the overall pipeline of the target video analysis application. In this work we propose a very efficient approach to capture the motion information within the video. Our method is based on a simple temporal and spatial derivation, which captures the changes between two consecutive frames. The proposed descriptor, Histograms of Motion Gradients (HMG), is validated on the UCF50 human action recognition dataset. Our HMG pipeline with several additional speed-ups is able to achieve real-time video processing and outperforms several well-known descriptors including descriptors based on the costly optical flow.
Tạp chí Khoa học Đo đạc và Bản đồ, 2013
Trên cơ sở nghiên cứu thành phần cấu tạo và các nguyên lý hoạt động của máy đo trọng lực hàng khô... more Trên cơ sở nghiên cứu thành phần cấu tạo và các nguyên lý hoạt động của máy đo trọng lực hàng không (TAGS Air III), tham khảo các dự án đo trọng lực hàng không đã thực hiện thành công tại các nước tiên tiến trên thế giới và khảo sát các điều kiện thực tế cùng các yên cầu của công tác đo trọng lực tại Việt Nam, chúng tôi đã xây dựng, đề xuất 1 quy trình vận hành bay khảo sát trọng lực ở Việt Nam như sau: (Xem hình 9)
Tạp chí Khoa học Đo đạc và Bản đồ, 2016
Trên thế giới các kết quả đo cao từ vệ tinh (Đo cao vệ tinh: Satellite Altimetry) như bề mặt biển... more Trên thế giới các kết quả đo cao từ vệ tinh (Đo cao vệ tinh: Satellite Altimetry) như bề mặt biển tự nhiên biển (MSS – Mean Sea Surface), mặt địa hình biển động lực trung bình so với Geoid (MDT – Mean Dynamic Topography), xác định trường trọng lực và Geoid trên biển v.v đã được ứng dụng nhiều trong nghiên cứu về Hải dương học, Khí tượng – Hải văn biển, Trắc địa và Địa vật lý. Ở Việt Nam việc nghiên cứu sử dụng mô hình MSS, MDT trong Trắc địa bản đồ cũng đã được một số tác giả khởi xướng từ vài năm trở lại đây. Trong bài báo này chúng tôi thử nghiệm sử dụng mô hình DNSC08MDT để tính trị đo sâu. Mô hình này được cung cấp chính thức tại thời điểm nghiên cứu, sau này đã có thêm các mô hình DTU10, DTU12 được cải tiến hơn (Dương Chí Công và nnk, 2015). Tại khu vực Bắc Trung Bộ (từ vĩ tuyến ~15° đến ~20°, từ bờ biển đến kinh độ 116°) với mô hình DNSC08MDT sau khi cải chính (về Geoid cục bộ Hòn Dấu trong hệ triều 0 và độ lệch hệ thống so với các trạm nghiệm triều) đã tính được trị đo sâu ch...
Environmental pollution (Barking, Essex : 1987), Jan 17, 2018
A massive and dense textured layer (ca. 35-50 cm thick) of hardpan was uncovered at the top layer... more A massive and dense textured layer (ca. 35-50 cm thick) of hardpan was uncovered at the top layer, which capped the unweathered sulfidic Cu-Pb-Zn tailings in depth and physically supported gravelly soil root zones sustaining native vegetation for more than a decade. For the purpose of understanding functional roles of the hardpan layer in the cover profile, the present study has characterized the microstructures of the hardpan profile at different depth compared with the tailings underneath the hardpans. A suit of microspectroscopic technologies was deployed to examine the hardpan samples, including field emission-scanning electron microscopy coupled with energy dispersive spectroscopy (FE-SEM-EDS), X-ray diffraction (XRD) and synchrotron-based X-ray absorption fine structure spectroscopy (XAFS). The XRD and Fe K-edge XAFS analysis revealed that pyrite in the tailings had been largely oxidised, while goethite and ferrihydrite had extensively accumulated in the hardpan. The percentag...
Journal of Ambient Intelligence and Humanized Computing, 2013
One of the key inputs for building energy and comfort management systems is detailed and precise ... more One of the key inputs for building energy and comfort management systems is detailed and precise office activity information, i.e., what is happening in each of the monitored areas (e.g., working place, meeting room, coffee corner, etc.). Such information can be used to control appliances for saving energy while still satisfying user comfort needs accordingly. We propose an activity recognition solution that effectively handles multiple-user, multiple-area situations, rapidly recognizing office activities as inputs for building energy and comfort management systems. The proposed solution is based on an ontological approach, using lowcost, binary, and wireless sensors. Through initial experiments, our prototype is able to recognize seven typical office activities (working at a desk with or without a PC, having a meeting, giving a presentation, having a coffee break, and presence/absence) at three office activity areas (working room, meeting room, and coffee corner) for two persons with average accuracy of more than 92%.
Energy and Buildings, 2013
Take-down policy If you believe that this document breaches copyright please contact us providing... more Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
2016 23rd International Conference on Pattern Recognition (ICPR), 2016
The encoding method is an important factor for an action recognition pipeline. One of the key poi... more The encoding method is an important factor for an action recognition pipeline. One of the key points for the encoding method is the assignment step. A very widely used super-vector encoding method is the vector of locally aggregated descriptors (VLAD), with very competitive results in many tasks. However, it considers only hard assignment and the criteria for the assignment is performed only from the features side, by looking for which visual word the features are voting. In this work we propose to encode deep features for videos using a double assignment VLAD (DA-VLAD). In addition to the traditional assignment for VLAD we perform a second assignment by taking into account the perspective from the codebook side: which are the nearest features to a visual word and not only which is the nearest centroid for the features as the standard assignment. Another important factor for the performance of an action recognition system is the feature extraction step. Recently, deep features obtained state-of-the-art results in many tasks, being also adopted for action recognition with competitive results over handcrafted features. This work includes a pipeline to extract local deep features for videos using any available network as a black box and we show competitive results including the case when the network was trained for another task or another dataset. Our DA-VLAD encoding method outperforms the traditional VLAD and we obtain state-of-the-art results on UCF50 dataset and competitive results on UCF101 dataset.
2016 14th International Workshop on Content-Based Multimedia Indexing (CBMI), 2016
Besides appearance information, the video contains temporal evolution, which represents an import... more Besides appearance information, the video contains temporal evolution, which represents an important and useful source of information about its content. Many video representation approaches are based on the motion information within the video. The common approach to extract the motion information is to compute the optical flow from the vertical and the horizontal temporal evolution of two consecutive frames. However, the computation of optical flow is very demanding in terms of computational cost, in many cases being the most significant processing step within the overall pipeline of the target video analysis application. In this work we propose a very efficient approach to capture the motion information within the video. Our method is based on a simple temporal and spatial derivation, which captures the changes between two consecutive frames. The proposed descriptor, Histograms of Motion Gradients (HMG), is validated on the UCF50 human action recognition dataset. Our HMG pipeline with several additional speed-ups is able to achieve real-time video processing and outperforms several well-known descriptors including descriptors based on the costly optical flow.
Tạp chí Khoa học Đo đạc và Bản đồ, 2013
Trên cơ sở nghiên cứu thành phần cấu tạo và các nguyên lý hoạt động của máy đo trọng lực hàng khô... more Trên cơ sở nghiên cứu thành phần cấu tạo và các nguyên lý hoạt động của máy đo trọng lực hàng không (TAGS Air III), tham khảo các dự án đo trọng lực hàng không đã thực hiện thành công tại các nước tiên tiến trên thế giới và khảo sát các điều kiện thực tế cùng các yên cầu của công tác đo trọng lực tại Việt Nam, chúng tôi đã xây dựng, đề xuất 1 quy trình vận hành bay khảo sát trọng lực ở Việt Nam như sau: (Xem hình 9)
Tạp chí Khoa học Đo đạc và Bản đồ, 2016
Trên thế giới các kết quả đo cao từ vệ tinh (Đo cao vệ tinh: Satellite Altimetry) như bề mặt biển... more Trên thế giới các kết quả đo cao từ vệ tinh (Đo cao vệ tinh: Satellite Altimetry) như bề mặt biển tự nhiên biển (MSS – Mean Sea Surface), mặt địa hình biển động lực trung bình so với Geoid (MDT – Mean Dynamic Topography), xác định trường trọng lực và Geoid trên biển v.v đã được ứng dụng nhiều trong nghiên cứu về Hải dương học, Khí tượng – Hải văn biển, Trắc địa và Địa vật lý. Ở Việt Nam việc nghiên cứu sử dụng mô hình MSS, MDT trong Trắc địa bản đồ cũng đã được một số tác giả khởi xướng từ vài năm trở lại đây. Trong bài báo này chúng tôi thử nghiệm sử dụng mô hình DNSC08MDT để tính trị đo sâu. Mô hình này được cung cấp chính thức tại thời điểm nghiên cứu, sau này đã có thêm các mô hình DTU10, DTU12 được cải tiến hơn (Dương Chí Công và nnk, 2015). Tại khu vực Bắc Trung Bộ (từ vĩ tuyến ~15° đến ~20°, từ bờ biển đến kinh độ 116°) với mô hình DNSC08MDT sau khi cải chính (về Geoid cục bộ Hòn Dấu trong hệ triều 0 và độ lệch hệ thống so với các trạm nghiệm triều) đã tính được trị đo sâu ch...
Environmental pollution (Barking, Essex : 1987), Jan 17, 2018
A massive and dense textured layer (ca. 35-50 cm thick) of hardpan was uncovered at the top layer... more A massive and dense textured layer (ca. 35-50 cm thick) of hardpan was uncovered at the top layer, which capped the unweathered sulfidic Cu-Pb-Zn tailings in depth and physically supported gravelly soil root zones sustaining native vegetation for more than a decade. For the purpose of understanding functional roles of the hardpan layer in the cover profile, the present study has characterized the microstructures of the hardpan profile at different depth compared with the tailings underneath the hardpans. A suit of microspectroscopic technologies was deployed to examine the hardpan samples, including field emission-scanning electron microscopy coupled with energy dispersive spectroscopy (FE-SEM-EDS), X-ray diffraction (XRD) and synchrotron-based X-ray absorption fine structure spectroscopy (XAFS). The XRD and Fe K-edge XAFS analysis revealed that pyrite in the tailings had been largely oxidised, while goethite and ferrihydrite had extensively accumulated in the hardpan. The percentag...