Video-Based Person Re-identification via 3D Convolutional Networks and Non-local Attention

Vehicle Speed Estimation Based on 3D ConvNets and Non-Local Blocks

Future Internet ◽

10.3390/fi11060123 ◽

2019 ◽

Vol 11 (6) ◽

pp. 123

Author(s):

Huanan Dong ◽

Ming Wen ◽

Zhouwang Yang

Keyword(s):

Optical Flow ◽

Camera Calibration ◽

Absolute Error ◽

Speed Estimation ◽

Vehicle Speed ◽

Convolutional Network ◽

Video Footage ◽

Convolutional Networks ◽

Calibration Methods ◽

Non Local

Vehicle speed estimation is an important problem in traffic surveillance. Many existing approaches to this problem are based on camera calibration. Two shortcomings exist for camera calibration-based methods. First, camera calibration methods are sensitive to the environment, which means the accuracy of the results are compromised in some situations where the environmental condition is not satisfied. Furthermore, camera calibration-based methods rely on vehicle trajectories acquired by a two-stage tracking and detection process. In an effort to overcome these shortcomings, we propose an alternate end-to-end method based on 3-dimensional convolutional networks (3D ConvNets). The proposed method bases average vehicle speed estimation on information from video footage. Our methods are characterized by the following three features. First, we use non-local blocks in our model to better capture spatial–temporal long-range dependency. Second, we use optical flow as an input in the model. Optical flow includes the information on the speed and direction of pixel motion in an image. Third, we construct a multi-scale convolutional network. This network extracts information on various characteristics of vehicles in motion. The proposed method showcases promising experimental results on commonly used dataset with mean absolute error (MAE) as 2.71 km/h and mean square error (MSE) as 14.62 .

Download Full-text

Densely Connected Graph Convolutional Networks for Graph-to-Sequence Learning

Transactions of the Association for Computational Linguistics ◽

10.1162/tacl_a_00269 ◽

2019 ◽

Vol 7 ◽

pp. 297-312 ◽

Cited By ~ 3

Author(s):

Zhijiang Guo ◽

Yan Zhang ◽

Zhiyang Teng ◽

Wei Lu

Keyword(s):

Sequence Learning ◽

Connected Graph ◽

Structural Information ◽

Text Generation ◽

Structural Representation ◽

Convolutional Network ◽

Neural Machine Translation ◽

Convolutional Networks ◽

Deep Architecture ◽

Non Local

We focus on graph-to-sequence learning, which can be framed as transducing graph structures to sequences for text generation. To capture structural information associated with graphs, we investigate the problem of encoding graphs using graph convolutional networks (GCNs). Unlike various existing approaches where shallow architectures were used for capturing local structural information only, we introduce a dense connection strategy, proposing a novel Densely Connected Graph Convolutional Network (DCGCN). Such a deep architecture is able to integrate both local and non-local features to learn a better structural representation of a graph. Our model outperforms the state-of-the-art neural models significantly on AMR-to-text generation and syntax-based neural machine translation.

Download Full-text

Local and Non-local Context Graph Convolutional Networks for Skeleton-Based Action Recognition

10.1007/978-3-030-86365-4_20 ◽

2021 ◽

pp. 243-254

Author(s):

Zikai Gao ◽

Yang Zhao ◽

Zhe Han ◽

Kang Wang ◽

Yong Dou

Keyword(s):

Action Recognition ◽

Local Context ◽

Convolutional Networks ◽

Non Local

Download Full-text

Pseudoinverse graph convolutional networks

Data Mining and Knowledge Discovery ◽

10.1007/s10618-021-00752-w ◽

2021 ◽

Author(s):

Dominik Alfke ◽

Martin Stoll

Keyword(s):

Real World ◽

Point Clouds ◽

Graph Laplacian ◽

Low Rank ◽

Low Rank Approximation ◽

Convolutional Networks ◽

3D Point Clouds ◽

Non Local ◽

Rank Approximation ◽

Real World Datasets

AbstractGraph Convolutional Networks (GCNs) have proven to be successful tools for semi-supervised classification on graph-based datasets. We propose a new GCN variant whose three-part filter space is targeted at dense graphs. Our examples include graphs generated from 3D point clouds with an increased focus on non-local information, as well as hypergraphs based on categorical data of real-world problems. These graphs differ from the common sparse benchmark graphs in terms of the spectral properties of their graph Laplacian. Most notably we observe large eigengaps, which are unfavorable for popular existing GCN architectures. Our method overcomes these issues by utilizing the pseudoinverse of the Laplacian. Another key ingredient is a low-rank approximation of the convolutional matrix, ensuring computational efficiency and increasing accuracy at the same time. We outline how the necessary eigeninformation can be computed efficiently in each applications and discuss the appropriate choice of the only metaparameter, the approximation rank. We finally showcase our method’s performance regarding runtime and accuracy in various experiments with real-world datasets.

Download Full-text

Toward High-Resolution Low-Voltage SEM

Proceedings, annual meeting, Electron Microscopy Society of America ◽

10.1017/s0424820100103152 ◽

1988 ◽

Vol 46 ◽

pp. 218-219

Author(s):

Zhifeng Shao

Keyword(s):

Low Voltage ◽

Spherical Aberration ◽

Chromatic Aberration ◽

Limiting Factor ◽

Operating Voltage ◽

Probe Radius ◽

Non Local ◽

Electron Microscopes ◽

Image Position ◽

Scanning Electron Microscopes

Recently, low voltage (≤5kV) scanning electron microscopes have become popular because of their unprecedented advantages, such as minimized charging effects and smaller specimen damage, etc. Perhaps the most important advantage of LVSEM is that they may be able to provide ultrahigh resolution since the interaction volume decreases when electron energy is reduced. It is obvious that no matter how low the operating voltage is, the resolution is always poorer than the probe radius. To achieve 10Å resolution at 5kV (including non-local effects), we would require a probe radius of 5∽6 Å. At low voltages, we can no longer ignore the effects of chromatic aberration because of the increased ratio δV/V. The 3rd order spherical aberration is another major limiting factor. The optimized aperture should be calculated as

Download Full-text

Chromatic aberration and low-voltage SEM

Proceedings, annual meeting, Electron Microscopy Society of America ◽

10.1017/s0424820100127293 ◽

1987 ◽

Vol 45 ◽

pp. 546-547

Author(s):

Zhifeng Shao ◽

A.V. Crewe

Keyword(s):

Electron Energy ◽

Low Voltage ◽

Spherical Aberration ◽

Chromatic Aberration ◽

Primary Electron ◽

Probe Size ◽

Non Local ◽

Optical Treatment ◽

Electron Microscopes ◽

Scanning Electron Microscopes

For scanning electron microscopes, it is plausible that by lowering the primary electron energy, one can decrease the volume of interaction and improve resolution. As shown by Crewe /1/, at V0 =5kV a 10Å resolution (including non-local effects) is possible. To achieve this, we would need a probe size about 5Å. However, at low voltages, the chromatic aberration becomes the major concern even for field emission sources. In this case, δV/V = 0.1 V/5kV = 2x10-5. As a rough estimate, it has been shown that /2/ the chromatic aberration δC should be less than ⅓ of δ0 the probe size determined by diffraction and spherical aberration in order to neglect its effect. But this did not take into account the distribution of electron energy. We will show that by using a wave optical treatment, the tolerance on the chromatic aberration is much larger than we expected.

Download Full-text