scholarly journals Lane following Learning Based on Semantic Segmentation with Chroma Key and Image Superposition

Electronics ◽  
2021 ◽  
Vol 10 (24) ◽  
pp. 3113
Author(s):  
Javier Corrochano ◽  
Juan M. Alonso-Weber ◽  
María Paz Sesmero ◽  
Araceli Sanchis
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Semantic Segmentation ◽  
Autonomous Driving ◽  
Imitation Learning ◽  
Small Scale ◽  
Car Model ◽  
Semantic Labeling ◽  
Indoor Scenarios ◽  
Chroma Key

There are various techniques to approach learning in autonomous driving; however, all of them suffer from some problems. In the case of imitation learning based on artificial neural networks, the system must learn to correctly identify the elements of the environment. In some cases, it takes a lot of effort to tag the images with the proper semantics. This is also relevant given the need to have very varied scenarios to train and to thus obtain an acceptable generalization capacity. In the present work, we propose a technique for automated semantic labeling. It is based on various learning phases using image superposition combining both scenarios with chromas and real indoor scenarios. This allows the generation of augmented datasets that facilitate the learning process. Further improvements by applying noise techniques are also studied. To carry out the validation, a small-scale car model is used that learns to automatically drive on a reduced circuit. A comparison with models that do not rely on semantic segmentation is also performed. The main contribution of our proposal is the possibility of generating datasets for real indoor scenarios with automatic semantic segmentation, without the need for endless human labeling tasks.

scholarly journals Semantic Image Segmentation with Deep Convolutional Neural Networks and Quick Shift

Symmetry ◽  
2020 ◽  
Vol 12 (3) ◽  
pp. 427 ◽  
Author(s):  
Sanxing Zhang ◽  
Zhenhuan Ma ◽  
Gang Zhang ◽  
Tao Lei ◽  
Rui Zhang ◽  
...  
Keyword(s):  
Neural Networks ◽  
Image Segmentation ◽  
Convolutional Neural Networks ◽  
Semantic Segmentation ◽  
Autonomous Driving ◽  
Input Image ◽  
Feature Representation ◽  
Segmentation Algorithm ◽  
Deep Convolutional Neural Networks ◽  
Semantic Image Segmentation

Semantic image segmentation, as one of the most popular tasks in computer vision, has been widely used in autonomous driving, robotics and other fields. Currently, deep convolutional neural networks (DCNNs) are driving major advances in semantic segmentation due to their powerful feature representation. However, DCNNs extract high-level feature representations by strided convolution, which makes it impossible to segment foreground objects precisely, especially when locating object boundaries. This paper presents a novel semantic segmentation algorithm with DeepLab v3+ and super-pixel segmentation algorithm-quick shift. DeepLab v3+ is employed to generate a class-indexed score map for the input image. Quick shift is applied to segment the input image into superpixels. Outputs of them are then fed into a class voting module to refine the semantic segmentation results. Extensive experiments on proposed semantic image segmentation are performed over PASCAL VOC 2012 dataset, and results that the proposed method can provide a more efficient solution.

Active Absorption of Random Waves in Wave Flume Using Artificial Neural Networks

2020 ◽  
Author(s):  
Áureo I. W. Ramos ◽  
Antonio C. Fernandes ◽  
Vanessa M. Thomaz
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Control Algorithm ◽  
Wave Generation ◽  
Reverse Direction ◽  
Small Scale ◽  
Wave Absorption ◽  
Wave Flume ◽  
Wave Channel ◽  
Artificial Neural

Abstract A wave flume is primarily intended to reproduce actual sea conditions in order to provide a reliable means of testing for small scale models. The realization of scaled tests is extremely important for the validation of a project in real scale, since, through the laws of similitude, such tests make it possible to predict the behavior of structures in the ocean as well as their performance during operation. This research aims to develop, test and validate an active control algorithm for wave absorption in a 2D wave channel — that is, when the waves propagate in only one direction — based on artificial neural networks (ANN). The ANN control algorithm relies on the linear wave theory and the principle of time reversal of wave propagation, i.e. the phenomenon of wave absorption corresponds to the wave generation when observed in the reverse direction of time. Through this principle, data from wave generation experiments, after proper manipulation, are used to train an ANN capable of generating the control signal used to move the wave generator device, this time as a wave absorber.

scholarly journals A method for mobile device positioning using a sensor network of BLE beacons, approximation of the RSSI value and artificial neural networks

2021 ◽  
Vol 45 (2) ◽  
pp. 277-285
Author(s):  
A.V. Astafiev ◽  
D.V. Titov ◽  
A.L. Zhiznyakov ◽  
A.A. Demidov
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Artificial Neural Networks ◽  
Sensor Network ◽  
Mobile Device ◽  
Experimental Studies ◽  
Test Sample ◽  
Small Scale ◽  
Approximation Function ◽  
Artificial Neural

The paper considers the development of a method for positioning a mobile device using a sensor network of BLE-beacons, the approximation of RSSI values and artificial neural networks. The aim of the work is to develop a method for positioning small-scale industrial mechanization equipment for building unmanned systems for product movement tracking. The work is divided into four main parts: data synthesis, signal filtering, selection of BLE beacons, translation of the RSSI values into a distance, and multilateration. A simplified Kalman filter is proposed for filtering the input signal to suppress Gaussian noise. A description of two approaches to translating the RSSI value into a distance is given: an exponential approximation function with a coefficient of determination of 0.6994 and an artificial feedforward neural network. A comparison of the results of these approaches is carried out on several test samples: a training one, a test sample at a known distance (0–50 meters) and a test sample at an unknown distance (60–100 meters). The artificial neural network is shown to perform better in all experiments, except for the test sample at a known distance (0–50 meters), for which the r.m.s. error is higher by 0.02 m 2 than that for the approximation function, which can be neglected. An algorithm for positioning a mobile device based on the multilateration method is proposed. Experimental studies of the developed method have shown that the positioning error does not exceed 0.9 meters in a 5×5.5 m room under monitoring. The positioning accuracy of a mobile device using the proposed method in the experiment is 40.9 % higher. Experimental studies are also conducted in a 58.4×4.5 m room, showing more accurate results compared to similar studies.

scholarly journals Flooding probability of urban area estimated by decision tree and artificial neural networks

2008 ◽  
Vol 10 (1) ◽  
pp. 57-67 ◽  
Author(s):  
Jeng-Chung Chen ◽  
Ching-Sung Shu ◽  
Shu-Kuang Ning ◽  
Ho-Wen Chen
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Small Scale ◽  
Urban Drainage ◽  
Rainfall Events ◽  
Second Stage ◽  
Downtown Area ◽  
Observation Network ◽  
Artificial Neural ◽  
Flooding Probability

Remote sensing, such as from satellite, has been recognized as useful for monitoring the changes in hydrology. In this study, we propose a way that is able to estimate flooding probability based on satellite data from the observation network of the World Meteorological Organization. Through a two-stage probability analysis, we can depict the area with high flooding potential in near-real time. In the first stage, decision trees offered a prompt and rough estimation of the flooding probability; in the second stage, artificial neural networks handle the rainfall forecast in a small-scale area. Case studies, simulating two rainfall events on 20 May 2004 and 11 July 2001, proved that our proposed method is promising for mitigating the flooding damage along urban drainage within the downtown area of Kaohsiung city.

scholarly journals Lightweight Convolutional Neural Networks with Model-Switching Architecture for Multi-Scenario Road Semantic Segmentation

2021 ◽  
Vol 11 (16) ◽  
pp. 7424
Author(s):  
Peng-Wei Lin ◽  
Chih-Ming Hsu
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Convolutional Neural Network ◽  
Semantic Segmentation ◽  
Autonomous Driving ◽  
Individual Model ◽  
Suppression Effect ◽  
Optimal Weights ◽  
Model Size ◽  
Multiple Scenarios

A convolutional neural network (CNN) that was trained using datasets for multiple scenarios was proposed to facilitate real-time road semantic segmentation for various scenarios encountered in autonomous driving. However, the CNN inhibited the mutual suppression effect between weights; thus, it did not perform as well as a network that was trained using a single scenario. To address this limitation, we used a model-switching architecture in the network and maintained the optimal weights of each individual model which required considerable space and computation. We, subsequently, incorporated a lightweight process into the model to reduce the model size and computational load. The experimental results indicated that the proposed lightweight CNN with a model-switching architecture outperformed and was faster than the conventional methods across multiple scenarios in road semantic segmentation.

scholarly journals DropConnect is effective in modeling uncertainty of Bayesian deep networks

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Aryan Mobiny ◽  
Pengyu Yuan ◽  
Supratik K. Moulik ◽  
Naveen Garg ◽  
Carol C. Wu ◽  
...  
Keyword(s):  
Neural Networks ◽  
Prediction Accuracy ◽  
State Of The Art ◽  
Computational Cost ◽  
Semantic Segmentation ◽  
Autonomous Driving ◽  
Learning Technology ◽  
Network Architectures ◽  
Uncertainty Estimates ◽  
Deep Networks

AbstractDeep neural networks (DNNs) have achieved state-of-the-art performance in many important domains, including medical diagnosis, security, and autonomous driving. In domains where safety is highly critical, an erroneous decision can result in serious consequences. While a perfect prediction accuracy is not always achievable, recent work on Bayesian deep networks shows that it is possible to know when DNNs are more likely to make mistakes. Knowing what DNNs do not know is desirable to increase the safety of deep learning technology in sensitive applications; Bayesian neural networks attempt to address this challenge. Traditional approaches are computationally intractable and do not scale well to large, complex neural network architectures. In this paper, we develop a theoretical framework to approximate Bayesian inference for DNNs by imposing a Bernoulli distribution on the model weights. This method called Monte Carlo DropConnect (MC-DropConnect) gives us a tool to represent the model uncertainty with little change in the overall model structure or computational cost. We extensively validate the proposed algorithm on multiple network architectures and datasets for classification and semantic segmentation tasks. We also propose new metrics to quantify uncertainty estimates. This enables an objective comparison between MC-DropConnect and prior approaches. Our empirical results demonstrate that the proposed framework yields significant improvement in both prediction accuracy and uncertainty estimation quality compared to the state of the art.

Sign in / Sign up

Export Citation Format

Share Document