scholarly journals Pedestrian Detection Based on Two-Stream UDN

2020 ◽  
Vol 10 (5) ◽  
pp. 1866 ◽  
Author(s):  
Wentong Wang ◽  
Lichun Wang ◽  
Xufei Ge ◽  
Jinghua Li ◽  
Baocai Yin
Keyword(s):  
Autonomous Vehicles ◽  
Pedestrian Detection ◽  
Activation Function ◽  
Practical Applications ◽  
Deep Network ◽  
The Road ◽  
Slowing Down ◽  
Deep Networks ◽  
The Difference ◽  
Computational Resources

Pedestrian detection is the core of the driver assistance system, which collects the road conditions through the radars or cameras on the vehicle, judges whether there is a pedestrian in front of the vehicle, supports decisions such as raising the alarm, automatically slowing down, or emergency stopping to keep pedestrians safe, and improves the security when the vehicle is moving. Suffering from weather, lighting, clothing, large pose variations, and occlusion, the current pedestrian detection still has a certain distance from the practical applications. In recent years, deep networks have shown excellent performance for image detection, recognition, and classification. Some researchers employed deep network for pedestrian detection and achieve great progress, but deep networks need huge computational resources, which make it difficult to put into practical applications. In real scenarios of autonomous vehicles, the computation ability is limited. Thus, the shallow networks such as UDN (Unified Deep Networks) is a better choice, since it performs well while consuming less computation resources. Based on UDN, this paper proposes a new deep network model named two-stream UDN, which augments another branch for solving traditional UDN’s indistinction of the difference between trees/telegraph poles and pedestrians. The new branch accepts the upper third part of the pedestrian image as input, and the partial image has less deformation, stable features, and more distinguished characters from other objects. For the proposed two-stream UDN, multi-input features including the HOG (Histogram of Oriented Gradients) feature, Sobel feature, color feature, and foreground regions extracted by GrabCut segmentation algorithms are fed. Compared with the original input of UDN, the multi-input features are more conducive for pedestrian detection, since the fused HOG features and significant objects are more significant for pedestrian detection. Two-stream UDN is trained through two steps. First, the two sub-networks are trained until converge; then, we fuse results of the two subnets as the final result and feed it back to the two subnets to fine tune network parameters synchronously. To improve the performance, Swish is adopted as the activation function to obtain a faster training speed, and positive samples are mirrored and rotated with small angles to make the positive and negative samples more balanced.

scholarly journals Pedestrian Detection Based on Two-Stream UDN

2020 ◽  
Author(s):  
Wentong Wang ◽  
Lichun Wang ◽  
Xufei Ge ◽  
Jinghua Li ◽  
Baocai Yin
Keyword(s):  
Autonomous Vehicle ◽  
Pedestrian Detection ◽  
Activation Function ◽  
Practical Applications ◽  
Deep Network ◽  
The Road ◽  
Slowing Down ◽  
Deep Networks ◽  
The Difference ◽  
Computational Resources

Pedestrian detection is the core of driver assistance system, which collects the road conditions through the radars or cameras on the vehicle, judges whether there is a pedestrian in front of the vehicle, supports decisions such as raising the alarm, automatically slowing down or emergency stopping to keep pedestrians safe, and improves the security when the vehicle is moving. Suffered from weather, lighting, clothing, large pose variations and occlusion, the current pedestrian detection still has a certain distance from the practical applications. In recent years, deep networks have shown excellent performance for image detection, recognition and classification. Some researchers employed deep network for pedestrian detection and achieve great progress, but deep networks need huge computational resources which make it difficult to put into practical applications. In real scenarios of autonomous vehicle, the computation ability is limited. Thus, the shallow networks such as UDN (Unified Deep Networks) is a better choice since it performs well on consuming less computation resources. Base on UDN, this paper proposes a new deep network model named as two-stream UDN, which augments another branch for solving traditional UDN’s indistinction of the difference between trees / telegraph poles and pedestrians. The new branch accepts the upper third part of the pedestrian image as input, and the partial image has less deformation, stable features and more distinguished characters from other objects. For the proposed two-stream UDN, multi-input features including HOG feature, Sobel feature, color feature and foreground regions extracted by GrabCut segmentation algorithms are fed. Compared with the original input of UDN, the multi-input features are more conducive for pedestrian detection since the fused HOG features and significant objects are more significant for pedestrian detection. Two-stream UDN is trained through two steps: First, the two sub-networks are trained until converge; then we fuse results of the two subnets as the final result and feed it back to the two subnets to fine tune network parameters synchronously. To improve the performance, Softplus is adopted as activation function to obtain faster training speed, and positive samples are mirrored and rotated with small angle to make positive and negative samples more balanced.

scholarly journals Effective Techniques for Pedestrian Detection in Smart Autonomous Vehicles

Webology ◽  
2021 ◽  
Vol 18 (05) ◽  
pp. 1176-1183
Author(s):  
Thylashri S ◽  
Manikandaprabu N ◽  
Jayakumar T ◽  
Vijayachitra S ◽  
Kiruthiga G
Keyword(s):  
Autonomous Vehicles ◽  
Autonomous Vehicle ◽  
Pedestrian Detection ◽  
Autonomous Driving ◽  
The Elderly ◽  
Driving Safety ◽  
Research Trend ◽  
Automatic Identification ◽  
Smart Vehicles ◽  
Slowing Down

Pedestrians are essential objects in computer vision. Pedestrian detection in images or videos plays an important role in many applications such as real-time monitoring, counting pedestrians at various events, detecting falls of the elderly, etc. It is formulated as a problem of the automatic identification and location of pedestrians in pictures or videos. In real images, the art of pedestrian detection is an important task for major applications such as video surveillance, autonomous driving systems, etc. Pedestrian detection is also an important feature of the autonomous vehicle driving system because it identifies pedestrians and minimizes accidents between vehicles and pedestrians. The research trend in the field of vehicle electronics and driving safety, vision-based pedestrian recognition technologies for smart vehicles have established themselves loudly or slowing down the vehicle. In general, the visual pedestrian detection progression capable of be busted down into three consecutive steps: pedestrian detection, pedestrian recognition, and pedestrian tracking. There is also visual pedestrian recognition in the vehicle. Finally, we study the challenges and evolution of research in the future.

A note on the applications of one primary function in deep neural networks

2021 ◽  
Author(s):  
Hengjie Chen ◽  
Zhong Li
Keyword(s):  
Neural Networks ◽  
Mathematical Knowledge ◽  
Deep Neural Networks ◽  
Activation Function ◽  
Continuous Functions ◽  
Middle Point ◽  
Deep Network ◽  
Rectified Linear Unit ◽  
The Mean ◽  
The Difference

By applying fundamental mathematical knowledge, this paper proves that the function [Formula: see text] is an integer no less than [Formula: see text] has the property that the difference between the function value of middle point of arbitrarily two adjacent equidistant distribution nodes on [Formula: see text] and the mean of function values of these two nodes is a constant depending only on the number of nodes if and only if [Formula: see text] By them, we establish an important result about deep neural networks that the function [Formula: see text] can be interpolated by a deep Rectified Linear Unit (ReLU) network with depth [Formula: see text] on the equidistant distribution nodes in interval [Formula: see text] and the error of approximation is [Formula: see text] Then based on the main result that has just been proven and the Chebyshev orthogonal polynomials, we construct a deep network and give the error estimate of approximation to polynomials and continuous functions, respectively. In addition, this paper constructs one deep network with local sparse connections, shared weights and activation function [Formula: see text] and discusses its density and complexity.

Analysis of Non-Linear Activation Functions for Classification Tasks Using Convolutional Neural Networks

2019 ◽  
Vol 12 (3) ◽  
pp. 156-161 ◽  
Author(s):  
Aman Dureja ◽  
Payal Pahwa
Keyword(s):  
Neural Networks ◽  
Deep Neural Networks ◽  
Activation Function ◽  
Primary Objective ◽  
Experimental Comparison ◽  
Activation Functions ◽  
Practical Applications ◽  
Network Activation ◽  
Non Linear ◽  
Hidden Layer

Background: In making the deep neural network, activation functions play an important role. But the choice of activation functions also affects the network in term of optimization and to retrieve the better results. Several activation functions have been introduced in machine learning for many practical applications. But which activation function should use at hidden layer of deep neural networks was not identified. Objective: The primary objective of this analysis was to describe which activation function must be used at hidden layers for deep neural networks to solve complex non-linear problems. Methods: The configuration for this comparative model was used by using the datasets of 2 classes (Cat/Dog). The number of Convolutional layer used in this network was 3 and the pooling layer was also introduced after each layer of CNN layer. The total of the dataset was divided into the two parts. The first 8000 images were mainly used for training the network and the next 2000 images were used for testing the network. Results: The experimental comparison was done by analyzing the network by taking different activation functions on each layer of CNN network. The validation error and accuracy on Cat/Dog dataset were analyzed using activation functions (ReLU, Tanh, Selu, PRelu, Elu) at number of hidden layers. Overall the Relu gave best performance with the validation loss at 25th Epoch 0.3912 and validation accuracy at 25th Epoch 0.8320. Conclusion: It is found that a CNN model with ReLU hidden layers (3 hidden layers here) gives best results and improve overall performance better in term of accuracy and speed. These advantages of ReLU in CNN at number of hidden layers are helpful to effectively and fast retrieval of images from the databases.

scholarly journals A Dynamic Distance Measure of Picture Fuzzy Sets and Its Application

Symmetry ◽  
2021 ◽  
Vol 13 (3) ◽  
pp. 436
Author(s):  
Ruirui Zhao ◽  
Minxia Luo ◽  
Shenggang Li
Keyword(s):  
Fuzzy Sets ◽  
Distance Measure ◽  
Distance Measures ◽  
Numerical Comparison ◽  
Mathematical Tool ◽  
Practical Applications ◽  
Fuzzy Point ◽  
Point Operator ◽  
The Difference ◽  
Picture Fuzzy Sets

Picture fuzzy sets, which are the extension of intuitionistic fuzzy sets, can deal with inconsistent information better in practical applications. A distance measure is an important mathematical tool to calculate the difference degree between picture fuzzy sets. Although some distance measures of picture fuzzy sets have been constructed, there are some unreasonable and counterintuitive cases. The main reason is that the existing distance measures do not or seldom consider the refusal degree of picture fuzzy sets. In order to solve these unreasonable and counterintuitive cases, in this paper, we propose a dynamic distance measure of picture fuzzy sets based on a picture fuzzy point operator. Through a numerical comparison and multi-criteria decision-making problems, we show that the proposed distance measure is reasonable and effective.

A robust and fast multispectral pedestrian detection deep network

2021 ◽  
pp. 106990
Author(s):  
Lu Ding ◽  
Yong Wang ◽  
Robert Laganière ◽  
Dan Huang ◽  
Xinbin Luo ◽  
...  
Keyword(s):  
Pedestrian Detection ◽  
Deep Network

scholarly journals CNN-Based Road-Surface Crack Detection Model That Responds to Brightness Changes

Electronics ◽  
2021 ◽  
Vol 10 (12) ◽  
pp. 1402
Author(s):  
Taehee Lee ◽  
Yeohwan Yoon ◽  
Chanjun Chun ◽  
Seungki Ryu
Keyword(s):  
Surface Crack ◽  
Autonomous Vehicles ◽  
Crack Detection ◽  
Model Performance ◽  
Semantic Segmentation ◽  
Road Surface ◽  
Image Brightness ◽  
The Road ◽  
Detection Model ◽  
Artificial Intelligence Models

Poor road-surface conditions pose a significant safety risk to vehicle operation, especially in the case of autonomous vehicles. Hence, maintenance of road surfaces will become even more important in the future. With the development of deep learning-based computer image processing technology, artificial intelligence models that evaluate road conditions are being actively researched. However, as the lighting conditions of the road surface vary depending on the weather, the model performance may degrade for an image whose brightness falls outside the range of the learned image, even for the same road. In this study, a semantic segmentation model with an autoencoder structure was developed for detecting road surface along with a CNN-based image preprocessing model. This setup ensures better road-surface crack detection by adjusting the image brightness before it is input into the road-crack detection model. When the preprocessing model was applied, the road-crack segmentation model exhibited consistent performance even under varying brightness values.

Deep learning based Blockchain solution for preserving privacy in future vehicles

2020 ◽  
pp. 1-14
Author(s):  
Varsha R ◽  
Meghna Manoj Nair ◽  
Siddharth M. Nair ◽  
Amit Kumar Tyagi
Keyword(s):  
Deep Learning ◽  
Autonomous Vehicles ◽  
Recommendation System ◽  
Privacy Preserving ◽  
Smart Devices ◽  
The Road ◽  
Blockchain Technology ◽  
Technological Advances ◽  
Previous Decade ◽  
Smart Things

The Internet of Things (smart things) is used in many sectors and applications due to recent technological advances. One of such application is in the transportation system, which is of primary use for the users to move from one place to another place. The smart devices which were embedded in vehicles are useful for the passengers to solve his/her query, wherein future vehicles will be fully automated to the advanced stage, i.e. future cars with driverless feature. These autonomous cars will help people a lot to reduce their time and increases their productivity in their respective (associated) business. In today’s generation and in the near future, privacy preserving and trust will be a major concern among users and autonomous vehicles and hence, this paper will be able to provide clarity for the same. Many attempts in previous decade have provided many efficient mechanisms, but they all work only with vehicles along with a driver. However, these mechanisms are not valid and useful for future vehicles. In this paper, we will use deep learning techniques for building trust using recommender systems and Blockchain technology for privacy preserving. We also maintain a certain level of trust via maintaining the highest level of privacy among users living in a particular environment. In this research, we developed a framework that could offer maximum trust or reliable communication to users over the road network. With this, we also preserve privacy of users during traveling, i.e., without revealing identity of respective users from Trusted Third Parties or even Location Based Service in reaching a destination. Thus, Deep Learning based Blockchain Solution (DLBS) is illustrated for providing an efficient recommendation system.

Determination of Resistance to Abrasive Wear. VIII. Influence of Softeners in Tests with the Akron-Croydon and Du Pont Machines

1949 ◽  
Vol 22 (1) ◽  
pp. 259-262
Author(s):  
J. F. Morley
Keyword(s):  
Unit Area ◽  
Published Data ◽  
The Road ◽  
Abrasive Surface ◽  
Abrasive Power ◽  
The Difference ◽  
On The Road ◽  
The Right ◽  
Abrasion Testing

Abstract These experiments indicate that softeners can influence abrasion resistance, as measured by laboratory machines, in some manner other than by altering the stress-strain properties of the rubber. One possible explanation is that the softener acts as a lubricant to the abrasive surface. Since this surface, in laboratory abrasion-testing machines, is relatively small, and comes repeatedly into contact with the rubber under test, it seems possible that it may become coated with a thin layer of softener that reduces its abrasive power. It would be interesting in this connection to try an abrasive machine in which a long continuous strip of abrasive material was used, no part of it being used more than once, so as to eliminate or minimize this lubricating effect. The fact that the effect of the softener is more pronounced on the du Pont than on the Akron-Croydon machine lends support to the lubrication hypothesis, because on the former machine the rate of wear per unit area of abrasive is much greater. Thus in the present tests the volume of rubber abraded per hr. per sq. cm. of abrasive surface ranges from 0.03 to 0.11 cc. on the du Pont machine and from 0.0035 to 0.0045 cc. on the Akron-Croydon machine. On the other hand, if the softener acts as a lubricant, it would be expected to reduce considerably the friction between the abrasive and the rubber and hence the energy used in dragging the rubber over the abrasive surface. The energy figures given in the right-hand columns of Tables 1 and 3, however, show that there is relatively little variation between the different rubbers. As a test of the lubrication hypothesis, it would be of interest to vary the conditions of test so that approximately the same amount of rubber per unit area of abrasive is abraded in a given time on both machines; this should show whether the phenomena observed under the present test conditions are due solely to the difference in rate of wear or to an inherent difference in the type of wear on the two machines. This could most conveniently be done by considerably reducing the load on the du Pont machine. In the original work on this machine the load was standardized at 8 pounds, but no figures are quoted to show how abrasion loss varies with the load. As an addition to the present investigation, it is proposed to examine the effect of this variation with special reference to rubbers containing various amounts and types of softener. Published data on the influence of softeners on the road wear of tire rubbers do not indicate anything like such large effects as are shown by the du Pont machine. This throws some doubt on the value of this machine for testing tire tread rubbers, a conclusion which is confirmed by information obtained from other workers.

Sign in / Sign up

Export Citation Format

Share Document