Article
Computer Science, Artificial Intelligence
Abbas Mohammadi, Kamal Jamshidi, Hamed Shahbazi, Mehran Rezaei
Summary: This paper proposes a low-cost method for reducing the computation resources required for deep convolutional neural networks (DCNN) in the steering control of self-driving cars, without sacrificing accuracy. The method introduces a feature density metric to filter out regions of input images that do not contain sufficient features, preventing unnecessary calculations. Compared to existing techniques, the proposed method significantly accelerates the training and inference phases.
Article
Computer Science, Artificial Intelligence
Yilmaz Kocak, Gulesen Ustundag Siray
Summary: This study aims to introduce new activation functions that combine the advantages of predefined activation functions to improve the performance of artificial neural networks. Researchers proposed several new activation functions and compared them with existing ones, using four different datasets to evaluate their performance.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Fernando Sevilla Martinez, Raul Parada, Jordi Casas-Roma
Summary: Autonomous driving and the machine learning models used for it have become increasingly important and complex, leading to a significant carbon footprint. This study introduces two different training approaches and compares their impact on carbon emissions. The new training method reduces training time by 38 times and decreases carbon emissions by 96%.
ADVANCED ENGINEERING INFORMATICS
(2023)
Article
Engineering, Civil
Ji-Seon Bang, Dong-Ok Won, Tae-Eui Kam, Seong-Whan Lee
Summary: Currently, there is a growing expectation for autonomous vehicles (AVs), but it will take at least a decade to develop fully autonomous vehicles without requiring human intervention. The risk of motion sickness (MS) is higher in AVs due to the sensory conflict theory, as passengers and drivers cannot predict the vehicle's movement path, leading to visual-perceptual dissonance. To maintain the driver's well-being, it is crucial to quickly predict and prevent MS through bio-signals and advanced driver assistance systems. This study used EEG data in real-world driving to predict MS, achieving an accuracy of 89.05% (+/- 5.76) using a normalized sample covariance matrix-based feature representation method and convolutional neural networks. This proposed model can serve as a valuable tool in resolving MS issues in AV environments.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Computer Science, Information Systems
Hajira Saleem, Faisal Riaz, Asadullah Shaikh, Khairan Rajab, Adel Rajab, Muhammad Akram, Mana Saleh Al Reshan
Summary: Deep learning techniques, especially convolutional neural networks (CNNs), have shown remarkable performance in solving vision-related problems, with a focus on architectural design and hyperparameter optimization globally. Metaheuristic algorithms have been proven superior for optimizing CNNs compared to manual-tuning. By applying the bat algorithm and particle swarm optimization algorithm, researchers have found improvements in the steering angle prediction problem in autonomous vehicles.
CMC-COMPUTERS MATERIALS & CONTINUA
(2022)
Article
Chemistry, Multidisciplinary
Ehtesham Hassan, V. L. Lekshmi
Summary: Despite significant research efforts, existing scene text detection methods are insufficient for real-life applications due to challenges posed by complex shapes, scale variations, and font properties in text segments. This paper proposes a novel scene text detector using a deep convolutional network that efficiently detects arbitrary oriented and complex-shaped text segments from natural scenes. The network design incorporates skip connections and text attention blocks based on efficient channel attention to capture complex text attributes at multiple scales. Extensive evaluations on various datasets show high detection F-scores.
APPLIED SCIENCES-BASEL
(2022)
Article
Computer Science, Artificial Intelligence
Tao Xu, Hongtao Wang, Guanyong Lu, Feng Wan, Mengqi Deng, Peng Qi, Anastasios Bezerianos, Cuntai Guan, Yu Sun
Summary: Due to the increasing number of fatal traffic accidents, there is a strong demand for more effective and convenient techniques for driving fatigue detection. This study proposes a unified framework called E-Key, which uses a convolutional neural network and attention structure to simultaneously perform personal identification and driving fatigue detection. The performance of the framework was assessed using EEG data collected from 31 healthy subjects, and it achieved the best performance in both personal identification (98.5%) and fatigue detection (97.8%) compared to other competitive models. The findings of this study show great potential for practical implementation in autonomous driving and car-sharing systems.
IEEE TRANSACTIONS ON AFFECTIVE COMPUTING
(2023)
Article
Engineering, Electrical & Electronic
Guanglong Du, Huijin Wang, Kang Su, Xueqian Wang, Shaohua Teng, Peter X. Liu
Summary: Driving fatigue is a significant factor in traffic accidents. Researchers have proposed a non-interference fatigue detection system that utilizes an electrocardiogram (ECG) acquisition device embedded in the steering wheel. By collecting the driver's ECG signals through their palm, the system can analyze tiredness after preprocessing. The system consists of a simulation generation module based on a cycle-generative adversarial network (CycleGAN) and a fatigue detection module based on a fuzzy convolution neural network (FCNN). Experimental results demonstrate the stability and accuracy of the proposed fatigue detection model.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Engineering, Multidisciplinary
Imtiaz Ul Hassan, Huma Zia, H. Sundus Fatima, Syed Adnan Yusuf, Muhammad Khurram
Summary: This paper introduces a lightweight CNN architecture for end-to-end learning in autonomous driving, which is 4 times lighter in parameters compared to Nvidia's PilotNet while achieving comparable results. Trained and evaluated using data from the CARLA simulator, the proposed model achieved a lower MSE than PilotNet.
MODELLING AND SIMULATION IN ENGINEERING
(2022)
Article
Multidisciplinary Sciences
Hong-Sik Kim, Inwhee Joe
Summary: Explainable Artificial Intelligence (XAI), a new trend in machine learning, aims to provide explanations for the outputs of machine learning models, especially in reliability-critical applications like self-driving cars. In this paper, the authors propose an XAI method based on computing and explaining the difference of output values in the last hidden layer of convolutional neural networks. The experimental results demonstrate its effectiveness in accurately identifying the parts needed to distinguish the category of images in self-driving cars.
Article
Computer Science, Information Systems
Mohsen Bakouri, Mohammed Alsehaimi, Husham Farouk Ismail, Khaled Alshareef, Ali Ganoun, Abdulrahman Alqahtani, Yousef Alharbi
Summary: Smart wheelchairs provide wheelchair users with self-dependence and improve their quality of life. In this study, a low-cost software and hardware method was designed and implemented to control a robotic wheelchair. An Android mobile app based on Flutter software was developed, along with a convolutional neural network and voice recognition model. The system achieved high accuracy and demonstrated good maneuverability performance for indoor and outdoor navigation.
Article
Engineering, Electrical & Electronic
Zhengchun Xie, Su Zhou, Miao Zheng, Fenglai Pei
Summary: A self-supervised algorithm based on deep learning is designed to estimate the depth of the driving scene. The algorithm utilizes the video information from a monocular camera to train the depth estimation network and pose estimation network. The algorithm addresses the issues of scale inconsistency and occlusion in the driving environment using view synthesis and scale consistency loss. The results show that the algorithm achieves high accuracy on the KITTI dataset.
SIGNAL IMAGE AND VIDEO PROCESSING
(2023)
Article
Construction & Building Technology
Fizza Hussain, Yasir Ali, Muhammad Irfan, Murtaza Ashraf, Shafeeq Ahmed
Summary: This study proposes a data-driven model based on Convolutional Neural Network (CNN) to predict the phase angle behavior of AC mixtures, which captures 90% of the variance in the test data. The model significantly improves upon other machine learning models and linear regression, providing a surrogate to tedious laboratory testing for transport agencies and practitioners.
CONSTRUCTION AND BUILDING MATERIALS
(2021)
Article
Neurosciences
Dongrui Gao, Xue Tang, Manqing Wan, Guo Huang, Yongqing Zhang
Summary: Driver fatigue detection is crucial for reducing accidents and improving traffic safety. This paper proposes a method based on log-Mel spectrogram and Convolution Recurrent Neural Network (CRNN) model using EEG signals to accurately detect driver fatigue. The method involves transforming the original EEG signal into spectrogram and using deep neural networks for feature extraction and classification. Experimental results show that the proposed method achieves high stability and outperforms existing methods.
FRONTIERS IN NEUROSCIENCE
(2023)
Article
Computer Science, Artificial Intelligence
Li Wang, Ziying Song, Xinyu Zhang, Chenfei Wang, Guoxin Zhang, Lei Zhu, Jun Li, Huaping Liu
Summary: Accurate 3D object detection from sparse LiDAR point cloud data is improved using a self-attention graph convolutional network (SAT-GCN). SAT-GCN utilizes GCN and self-attention to enhance semantic representations and improve detection performance. The proposed method achieves significant improvements on popular 3D object detection benchmarks, demonstrating its effectiveness in enhancing the detection accuracy of point cloud data.
KNOWLEDGE-BASED SYSTEMS
(2023)