Proceedings of the 5th International Conference on Computing and Data Science
Roman Bauer, University of Surrey
Marwan Omar, Illinois Institute of Technology
Alan Wang, University of Auckland
In recent years, with the quantum leap in deep learning, self-driving vehicle as one of its applications has been gaining tremendously increasing popularity as well as making a multitude of achievements. Object detection, which has made significant contribution to driver-less vehicle, has had been applied to a tremendously wide range of fields. However, reports relevant to automatic vehicle stating that accidents are caused by Automatic driving technology, present problems pointing out that existing target detection algorithms, which are already fairly well reliable, can probably be interfered by adverse conditions such as high temperature, raise dust and transmission loss, and be not capable of providing precise output. This paper recaps on these previous classic algorithms, and their large-scale application domain. Meanwhile, this paper presents improvements focusing on enhancing the robustness of these algorithms to overcome these problems caused by adverse conditions and improve the accuracy. Thus these improvements could augment the security of these driver-less vehicles, and eventually reduce traffic accident mortality relative to self-driving vehicles and safeguard road safety, and may potentially benefit to further research.
Motion Capture technology refers to the technology of recording and processing the actions of people or other objects, and generating corresponding virtual asset action animations. It has been widely used in our daily life. Virtual characters like Hulk, Gollm, and Avatar were all built based on the technology. It can also be applied to sports analysis, human biomechanics, automotive, virtual reality, etc, showing its huge application prospects. Specifically, 3D Motion Capture Market was forecast to reach $270.9 million by 2026 suggested by industry ARC. To help readers get familiar with the field without much background knowledge, this paper reviews the origin of the motion capture technology, the way motion capture technology was realized and the mainstream methodologies of capturing motion. This paper provides the reader with some existing applications of motion capture technology in different fields so that the reader can understand the importance of the technique. Readers can also get some insight into what future studies may be focused on and if there is any promotion of the motion-capturing methods.
Three-dimensional objects are usually represented by point cloud based on lidar reflection of the sensors. However, the point clouds are commonly sparse since the lidar reflection is restricted by the location where the machine scans the objects around it. Commonly, the point cloud data that we use in autonomous driving are basically concentrated on one side or two sides and can seldom depict the whole image of all the objects. In this paper, we propose methods based on symmetry to diminish this intrinsic problem. Our work uses CenterPoint as our backbone and we do some fine-tune on it to make the data augmented. Moreover, we use FutureDet as the detector and the predictor to see whether the results of the methods fit the design. We obtain the information of CenterPoint from the detector and use the position information of the center point to do symmetry so as to make the points exist on the other side. We are trying to address the issue by comparing the results of metrics from the FutureDet and the fine-tune model on nuScenes dataset.
With many gaming AI being developed and being able to defeat top human players in recent years, AI has once again become the hottest topic in research and even in our daily life. This paper also researches on gaming AI and chess. Instead of using deep learning and the Monte Carlo Search algorithm, this paper focuses on the opening only with multi-armed bandit algorithms to find the best moves and opening. Specifically, the method used in this paper is epsilon greedy and Thompson sampling. The dataset used in this paper is from Kaggle. This paper considers each move as a set of choices one needs to make and considers the big picture as a multi-armed bandit problem. This paper aims to develop a relative best strategy to counter those opening or make changes to a disadvantaged situation.
Explore-then-commit (ETC) algorithm is a widely used algorithm in bandit problems, which are used to identify the optimal choice among a series of choices that yield random outcomes. The ETC algorithm is adapted from A/B testing, a popular procedure in decision-making process. This paper explores the multi-armed bandit problem and some related algorithms to tackle the multi-armed bandit problem. In particular, this paper focuses on the explore-then-commit (ETC) algorithm, a simple algorithm that has an exploration phase, and then commits the best action. To evaluate the performance of ETC, a variety of settings is made in the experiment, such as the number of arms and input parameter m, i.e., how many times each arm is pulled in the exploration phase. The result shows that the average cumulative regret increases when the number of arms gets larger. With the increase of parameter m, the cumulative regret decreases in the beginning, until reaching the minimum value, and then starts increasing. The purpose of this paper is to empirically evaluate the performance of the ETC algorithm and investigate the relationships between the parameter settings and the overall performance of the algorithm.
In the past decades, tons of data have been generated every single day as people increasingly rely on electronic products and networks in their lives. The developments in techniques and storage capabilities provide a fundamental condition for analyzing those huge-volumed data. Each captured feature represents a dimension of the data. So far, high dimensional data analysis has become a challenging task in various study fields. Redundant and irrelevant features can be removed by using different dimensionality reduction techniques. Effective information extraction can be achieved by using the proper method. This paper reviews the most widely used dimensionality reduction techniques and their application fields. It can be found that though DRTs have been successfully applied to many areas (i.e., image, audio/video data, biomedical), DRTs still need to be improved and developed to achieve better classification and prediction accuracy. Inter-method combinations will remain the focus of research in the future. Computation time and cost may not be a limitation anymore as the computation power of the computer is still developing. So, the development of the algorithm is becoming particularly important. This study provides a brief introduction to widely used DRTs and their variants, it will be helpful for understanding HDD analysis, and more DRTs will be researched in future work.
Crop disease detection is an important factor in agricultural production. Traditional object detection methods can't effectively screen key features, resulting in weak crop disease control in many countries. In recent years, several convolutional neural networks for object detection have been proposed, which makes it possible to apply computer vision to crop disease identification through deep learning. YOLOv5 is an advanced object detection network, which can extract key features and use human visual attention mechanism as a reference. This paper would evaluate the performance of four pre-trained models of YOLOv5 in object detection of crop diseases. And transfer learning was used to train the corresponding dataset. The experiment results have showed that the F1 values of the four models all reached above 0.93, and the Yolov5x got the best result, which achieved 0.963. Furthermore, the detection accuracy of the four models has reached more than 98%. This shows that the YOLOv5 series network models have great application prospects in the identification of crop diseases. In the near future, the object detection model can be applied to various mobile devices, even unmanned aerial vehicles, which would play a significant role in crop disease prevention.
The spring inverted pendulum model was used to analyze the motion of the biped robots under different motion states and the effective model parameters were obtained. In this paper, the biped robot was simplified to an inverted pendulum model and the kinematics equations of the robot under different motion states were obtained. The kinematics equations were solved by MATLAB, and the analysis was used to get the right sports condition during motion process. The corresponding simulation of the inverted pendulum model was carried out based on MATLAB, a series of data such as the trajectory of mass point of the robot and the variation of the ground reaction force during motion process were obtained. By analyzing the results, the model parameters and conditions of the biped robots to keep stable in different motion states were obtained.
With the rapid growth of the internet today, cloud storage services have had an impact on numerous internet users worldwide. Internet users, in order to avoid potential risk of data loss, can outsource their local data to remote cloud servers instead of using local media. However, this kind of cloud storage services is sometimes unreliable enough and the data security of users cannot be guaranteed. In this case, data auditing mechanisms for cloud storage are studied to avoid the destruction of user data. The traditional audit schemes are based on provable data possession mechanisms, while the third-party organization relies on the recent blockchain-based methods, conventional single-cloud storage system and recently emerging multi-cloud storage models. Some audit schemes not only provide auditing capability, but also have some other ancillary functions to ensure applicability for different potential malevolent situations. Gathered correlated studies about data auditing schemes of cloud service are analyzed and classified in this paper, based on research objects and research methods in order to seek probable innovation points. We also analyze experiments in their papers. Toward the finish of this review, we sum up all the examination, list the bearings not investigated, and give a few creative spots.
A language is a valuable tool for human development and progress, and it is also an important medium for human beings to transmit information and express emotions. Language signals are ubiquitous, and it is an indispensable part of human life. This article will take the analysis of language as the starting point, combined with the relevant content of computer deep learning, and summarize various methods of language emotion recognition based on a convolutional neural network. In recent years, with the gradual intelligentization of computers, more in-depth discoveries and research have been made on language emotion research. In the deep neural network sector, most of the models used are CNN, LSTM, MO-LSTM models, and this paper aims to propose a new CLDNN (CONVOLUTIONAL, LONG SHORT-TERM) that integrates CNN, DNN, and LSTM into the same network. MEMORY, FULL CONNECTED DEEP NEURAL NETWORKS) model, compare with it, and summarize the advantages and disadvantages of CLDNN.
As a naturally porous medium, rocks have a very strong heterogeneity in their internal pore structure, which has an important impact on the mechanical and chemical properties of rocks, etc. Therefore, it is very important to quantitatively characterise the heterogeneity of the pore structure of rocks. Fractal dimension has long been recognised as an effective means of characterising the heterogeneity of the pore structure of porous media and is widely used in oil and gas exploitation, construction materials, mining, and water engineering. There are different methods to calculate the fractal dimension. To verify the consistency between different methods, this paper compares the results of fractal analysis using the mercury intrusion method and MATLAB image fractal analysis using three rock samples with large differences in porosity and calculates the fractal dimension in three different ways on the basis of the mercury intrusion method. The results demonstrate that the fractal dimension of the mercury intrusion method and the box counting dimension of the image analysis obtained by the three methods, although slightly different in numerical value, are consistent in their numerical relationship, i.e., they all conform to the rule that the stronger the non-homogeneity of the pore throat, the larger the fractal dimension. The results of this paper show that fractal dimension is indeed an effective means of characterising rock homogeneity.
This paper aims to explore the relationship between focus and accent, namely, whether the focus of a sentence possesses some acoustic features of accent. Previous studies suggest that focused words depict different patterns of acoustic prominence embedded in accent. According to previous evidence found in English, pitch and duration are two crucial factors contributing to differentiating focused and unfocused words. Therefore, pitch and duration are the main two factor considered in this study. Studies in aspect of the relationship between focus and accent and their realization in the phonetic level in natural speech are insufficient. Therefore, this study aims to provide support for the connection between focus and accent in natural speech in Mandarin. Based on 24 sentences from three native Mandarin speakers selected from a Chinese corpus, THCHS-30, published by Tsinghua University, this study extracts the narrow-focused words, broad-focused words, and the corresponding unfocused-words as the main subjects and analyses their mutual relationships. Results support the symmetric relationship between focus and accent. Focused words show phonetic features of accent and words with no obvious phonetic prominence of accent are unlikely to be considered as focus. Focus triggers a decrease in F0 values and an increase in duration of focused words. Linear mixed effects tests in R suggest that a major effect of narrow and broad focus in causing F0 and duration changes has not been deducted, but narrow focus conditions can result in duration proportion increase of narrow focused words compared to broad focused ones.
With advances in mobile technology and mobile Internet applications, smart mobile devices, such as smartphones and tablets, have become increasingly popular, and the number of Internet users worldwide continues to grow. In the Internet era, the amount of data is growing exponentially and companies must be able to harness the value of the vast amount of data. Data platforms must integrate massive amounts of data collection, storage, computation and analysis to meet these opportunities and challenges. In this study, the log data of Internet users browsing websites are analyzed and the technologies used in the platform are briefly described. Finally, a draft platform for analyzing offline Internet user behavior data is proposed, taking into account the current common needs of different industries, while incorporating some innovations. Three modules are designed and implemented: data collection, data warehouse and data visualization. The user's data is mainly collected by the data collection module. The data warehouse is mainly responsible for cleaning, modeling and analyzing the data. As part of the data visualization module, the result data from the ADS layer is used as a template to create tables in MySQL, export the results to MySQL periodically using the Sqoop tool, and visualize the data using the data visualization tool. With Flume, Kafka and Sqoop, HDFS is used as the data storage framework, Hive is used as the storage tool, and Spark is used as the Hive computation engine to build the platform in a large context to analyze Internet user behavior.
Today's buildings usually rely on various firefighting equipment for warning, often neglecting the importance of safety inspections. Most fires occur due to the usual weak awareness of fire prevention. If regular and strict safety inspections are carried out according to the regulations of fire departments will greatly reduce the occurrence of fires. Safety inspections in each unit include fire supervisors from the government and inspectors from relevant maintenance agencies, and these inspectors usually use paper reports to record the safety inspection work. This traditional practice leaves hidden problems for unresolved and important issues, including data integrity and tamperability, which will directly lead to difficulties in tracing responsibilities based on these data after an accident. Therefore, this paper proposes a de-neutralized, blockchain-based traceability system for a fire safety inspection. Blockchain technology has the characteristics of traceability, decentralization, consensus trust, and hard to tamper, which can be combined with this system. The system's core mainly adopts the Hyperledger Fabric consortium blockchain framework, configured to build a blockchain network. The network is organized with three organizations: government, unit, and maintenance agency, and smart contracts are written to process safety inspection data, thus guaranteeing the sharing of fire inspection information as well as traceability.
With the development of sensory technologies, the inertial measurement unit (IMU) has been emerging widely. In this paper, the IMU sensor is used to facilitate the versatile applications of robotic manipulators. The Kalman filter fuses the raw measurements of an IMU mounted on human hands to obtain the real-time attitude information. The contribution is to manipulate the robotic arm dexterously by the motion guidance of human hands, to achieve human-robot interaction. The experimental results evaluate the attitude preciseness and associated manipulator’s locomotion. In the future work, the proposed interaction method will be applied in manipulation tasks such as grasping, inspection and so on.
The fast development in American multinational technology companies has attracted both professional and new investors to buy the stocks. However, the price of these companies are unstable and therefore hard to be predicted. The focus of this article is to use AI and deep learning algorithms to find a pattern of the stock price. Long Short-Term Memory Algorithm (LSTM) is the main algorithm used to predict the trend, and other methods including Autoregressive integrated moving average (ARIMA), Seasonal autoregressive integrated moving average (SARIMA), and prophet are also discussed in this piece.
With the increasing use of heating, ventilating, and air conditioning (HVAC) systems nowadays, their energy consumption is receiving more attention. The study begins by trying available anomaly detection techniques, including KNN, COF, and isolated forests. The comparison reveals that these methods disregard some linear correlation results. Then, the pattern is summarized by analyzing data from 100 HVAC-equipped rooms. Next, the study uses correlation analysis and neural networks to identify abnormal HVAC data. Finally, it concludes by analyzing the factors that lead to the anomalies.
Recent research has shown that self-attention mechanisms improve Sequential Recommender Systems (SRS) by capturing sequential associations with the interactions. Nevertheless, existing work still needs to address two critical limitations. Firstly, the behavior of users in the original sequences contains various preference signals that are implicit and noisy and hard to reflect the user’s intentions fully. As a result, it would deteriorate the representation of their true intentions to model all interactions. Secondly, most models only model single-scale interaction sequences and ignore the multi-scale feature relationships of the sequences. In order to address these limitations, the paper proposes MFTSRec (Multi-scale Filter Enhanced Transformer Sequential Recommender), which can weaken those interactions irrelevant to the users’ intentions from their implicit feedback and adaptively focus on the user’s multi-scale intentions. Besides, this paper also does extensive experiments on four benchmark datasets and further demonstrates the effectiveness and robustness of MFTSRec compared to the state-of-the-art model.
Deep learning is widely used in various fields, the article proposed a deep learning method to predict house prices through different characteristics of real estate, establish a prediction model, and carry out simulation experiments. First, extracting data from property transactions records, it is difficult to directly input the raw-data into the deep learning model, and there may be overfitting in the model training, so data will be pretreated. Second, Fully-Connected Neural Network is used to model different features’ influences to price. The sample data will be randomly divided into a training set and a validation set, of which 70% of the samples are used for building and training the model, and the remaining 30% are used for model accuracy verification. Experimental results show that the model can achieve a high accuracy in predicting houses price. The model can be used as a reference for the evaluation of housing prices.
The purpose of the paper is to tackle the classification problem of 3D point cloud data in domain generalization: how to develop a generalized feature representation for an unseen target domain by utilizing sub-field of numerous seen source domain(s). We present a novel methodology based on both adversarial training to learn a generalized feature representations across subdomains in domain adaptation called 3D-AA. We specifically expand adversarial autoencoders by applying the Maximum Mean Discrepancy (MMD) measure to align the distributions across several subdomains, and then matching the aligned distribution to any given prior distribution via adversarial feature learning. In this manner, the learned 3D feature representation is supposed to be universal to the observed source domains due to the MMD regularization and is expected to generalize well on the target domain due to the addition of the prior distribution. We applied an algorithm to train two different 3D point cloud source domains with our framework. The combination of multiple loss functions on 3D point cloud domain generalization task show that our applied algorithm performs better and learn more generalized features for the target domain than the source-only algorithm which only utilized the MMD measurement.