{"title":"Video Tampering Detection based on High-Frequency Features using Machine Learning","authors":"Hiroki Ueda, Hyunho Kang, Keiichi Iwamura","doi":"10.1145/3442536.3442540","DOIUrl":"https://doi.org/10.1145/3442536.3442540","url":null,"abstract":"In recent times, security cameras can be found installed in various places such as lobbies of buildings, urban areas, and drive recorders in automobiles. Similarly, smartphones and drive recorders have become a part of our everyday lives. However, in recent years, advances are being made in technology that enables images and videos to be edited easily. For example, editing software, such as Photoshop, After Effects, and Premiere Pro, allow anyone to edit videos and images easily. Thus, it is easy for people to take pictures or videos using smartphones or drive recorders and tamper with them. When an accident or crime occurs, videos or images from security cameras can serve as legal evidence for the genuineness of the background or object. Therefore, it is necessary to investigate the possibility of tampering with the video contents for malicious purposes. In this study, we propose a method for detecting video tampering based on high-frequency features using machine learning. Artifacts of tampered videos can be detected in consecutive frames through video forgery detection. In addition, although machine learning in this field is limited by its inability to acquire sufficient manipulated data, we demonstrate that this can be improved.","PeriodicalId":353230,"journal":{"name":"Proceedings of the 2020 3rd Artificial Intelligence and Cloud Computing Conference","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126953944","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Towards Diagnosis of Carpal Tunnel Syndrome Using Machine Learning","authors":"Yuan Wei, Wei Zhang, F. Gu","doi":"10.1145/3442536.3442549","DOIUrl":"https://doi.org/10.1145/3442536.3442549","url":null,"abstract":"Carpal Tunnel Syndrome (CTS) is the most common peripheral neuropathy affecting the hand function. Although the most complains from patients with CTS are fine motor control failures in daily manual activities, parameters of hand functional control have not been considered neither in current diagnostic nor evaluation process. In addition, CTS has been identified as an occupational disease. Over 50% of reported CTS cases are work related. However, early screening protocols of CTS at a preliminary stage are absent, and thus unable to prevent further complications, especially for high-risk populations who can advance their CTS stage on daily work basis. In the current protocol, we aim to identify important parameters of hand functional control that are indicative of CTS clinical occurrence and severity stages. Based on designed experiments during hand grasping, we performed machine learning classifiers to detect, filter and subtract important biomarkers or groups of biomarkers that dominantly classify the CTS hand and its severity. The identified biomarkers not only provide a high potential of a paradigm shift in CTS management, but also are able to shed light on hand functional evaluations associated with this neuropathy. In this paper, we adopt one of machine learning approaches, random forests, to the raw experimental hand function gripping data to identify the most important biomarkers for CTS. The experimental results show the effectiveness of the proposed work.","PeriodicalId":353230,"journal":{"name":"Proceedings of the 2020 3rd Artificial Intelligence and Cloud Computing Conference","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122133399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"How Do We Predict Stock Returns in the Cross-Section with Machine Learning?","authors":"Masaya Abe, Kei Nakagawa","doi":"10.1145/3442536.3442547","DOIUrl":"https://doi.org/10.1145/3442536.3442547","url":null,"abstract":"Stock return prediction is one of the most important themes for investors. Until now, there are many studies for the application of machine learning methods to predict stock returns in the cross-section. However, those studies focus only on differences in machine learning methods. This study investigates how the difference in problem settings such as problem definition and data preprocessing affects the performance of stock return prediction. Our results show that the performance varies depending on problem settings regardless of the prediction models. These findings indicate that not only the prediction models but also the problem settings are important for stock return prediction.","PeriodicalId":353230,"journal":{"name":"Proceedings of the 2020 3rd Artificial Intelligence and Cloud Computing Conference","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128419946","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Video Object Detection Based on Deformable Convolution","authors":"Yanjun Liang, Zhiming Wang","doi":"10.1145/3442536.3442537","DOIUrl":"https://doi.org/10.1145/3442536.3442537","url":null,"abstract":"Video object detection is one of the important research directions in the field of computer vision with applications in various domains, e.g. public security, traffic management, etc. Nevertheless, it is very challenging to extend the image-based object detection method to video object detection. In the existing methods, the feature information of multiple adjacent frames is aggregated into the current frame to improve the detection result. However, these methods only sample the information of adjacent frames on the sampling points obtained by prior knowledge or by calculating explicit motion references. Therefore, the feature information obtained by this sampling method is often insufficient. Furthermore, if the selected sampling points are inaccurate, the original feature information may be lost. In this paper, we use modulation-based deformable convolution for feature sampling, and the sampling positions and weight are generated through additional convolutional layers. This not only enriches the sampling information but also minimizes the intervention of prior knowledge. According to experiments on the Image VID dataset, the proposed method significantly improve detection accuracy compared with existing methods.","PeriodicalId":353230,"journal":{"name":"Proceedings of the 2020 3rd Artificial Intelligence and Cloud Computing Conference","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128474506","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"The Walking Robot Based on Mechanical Connecting-rod","authors":"Jian-Bing Zhang, Xiao-Ping Wang","doi":"10.1145/3442536.3442545","DOIUrl":"https://doi.org/10.1145/3442536.3442545","url":null,"abstract":"In recent years, China's investment in space technology and natural environment has been increasing. In recent years, China's investment in space technology and natural environment has been increasing. First of all, it must be the bionic robot with strong adaptability to explore the surfaces of the moon, Mars and other planets under an uncertain environment. Compared with wheeled or tracked robot, the standpoints of the legged robot are discrete points, which can select optimal supporting spots on the ground, so the legged robot is better, especially the legged robot with eight legs. And the walking robot based on connecting-rod has been widely used because of its simple structure and strong load capacity. Secondly, due to the wide use of computer in society as well as the fast development of integrated circuit, SCM is playing a greater role. It has a lot of functions, can work steadily, and can be used conveniently but with small volume and low price so many of them are applied in the control systems, but more applications in the automatic control system, data acquisition, military products as well as appliances and other fields. And according to the specific hardware structure and the object, using the software makes it more perfect. In this paper, we designed a walking robot adopted the structure of connecting-rot, and take the SCM LM298 as the core of controlling. The robot takes its operation by using the integrated chips of encoding and decoding. This walking robot has realized many functions such as walking, turning back, left and right.","PeriodicalId":353230,"journal":{"name":"Proceedings of the 2020 3rd Artificial Intelligence and Cloud Computing Conference","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122331243","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Using Twitter Social Media for Depression Detection in the Canadian Population","authors":"Ruba Skaik, D. Inkpen","doi":"10.1145/3442536.3442553","DOIUrl":"https://doi.org/10.1145/3442536.3442553","url":null,"abstract":"Depression is a serious public health problem, and an economic burden for the society. Therefore, identifying individuals with depression and providing the support for the people in need is a crucial step for creating a healthier environment. In this research, we explore the task of detecting signs of depression from tweets where people often express their feelings, thoughts, interests, and opinions. We utilize personal narratives collected users with self-reported depression to build a model suitable for predicting depression in a sample of Twitter users that is representative for the Canadian population. The training dataset contains 1,402 users who self-reported depression diagnoses during 2016. Using classical machine learning techniques, we achieved 0.961 F1-score using 10-fold cross validation. We used the CLPsych 2015 dataset as a test dataset, and we reached 0.898 F1-score using deep learning models. Then, we applied the same model on a population sample and obtained a result in line with the depression statistics reported by Statistics Canada for 2015.","PeriodicalId":353230,"journal":{"name":"Proceedings of the 2020 3rd Artificial Intelligence and Cloud Computing Conference","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131794104","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Development of Robotic Quiz Games for Self-Regulated Learning of Primary School Children","authors":"Tingsheng Weng, Chien-Kuo Li, Meng-Hui Hsu","doi":"10.1145/3442536.3442546","DOIUrl":"https://doi.org/10.1145/3442536.3442546","url":null,"abstract":"The progressive development of information technology has provided multiple learning modes. The rich content and innovative applications available allow pupils to improve their skills through self-regulated learning (SRL), which has become an important education goal. Intelligent robots can be used in a wide range of applications, from programmed movements for learning activities, to the combination of artificial intelligence and sensor technology for human life and education. A robot's dynamic and interesting interface is more suitable for children's self-regulated learning. This study used a Zenbo robot as the development tool and Zenbo Scratch platform programming to develop an AI robot math quiz game for primary school students. Two elementary school math teachers, and a parent and a 5th grade primary school student were involved in the development of the game. This study used the parent's and student's continuous interaction with the robot to adjust the code and achieve the best human-computer interaction in robotic mathematics problem solving. Moreover, this study developed a companion robot for a math quiz game, which can be used for reviewing what has been learned in class. The robot can be used for self-regulated learning by young children to increase student learning outcome.","PeriodicalId":353230,"journal":{"name":"Proceedings of the 2020 3rd Artificial Intelligence and Cloud Computing Conference","volume":"334 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114526933","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"CNN as a feature extractor in gaze recognition","authors":"A. Govindaswamy, E. Montague, D. Raicu, J. Furst","doi":"10.1145/3442536.3442542","DOIUrl":"https://doi.org/10.1145/3442536.3442542","url":null,"abstract":"In this paper, we employ a Convolutional Neural Network (CNN) in predicting physician gaze. This paper focuses on two aspects – one comparison between hand-crafted features and CNN-based learned features, and two in investigating the impact of fully-connected layers in an end-to-end CNN model. The pre-trained CNN model based on VGG16 through transfer learning is used as a feature extractor and a K-Nearest Neighbor and a Random Forest (RF) algorithm were used as the classifier of physician gaze. The CNN-RF and CNN–K-NN models were compared with the traditional end-to-end CNN model and through a series of experiments and statistical tests of significance, we show that the power of CNN comes from the features extraction part and that the fully connected layers of the CNN have comparable performance to the random forest and the k-NN classifiers. We also show that the CNN-based learned features provide substantial distinguishable power in classifying physician gaze.","PeriodicalId":353230,"journal":{"name":"Proceedings of the 2020 3rd Artificial Intelligence and Cloud Computing Conference","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122243770","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Detecting Discrete Cosine Transform-Based Digital Watermarking Insertion Area Using Deep Learning","authors":"Sayoko Kakikura, Hyunho Kang, Keiichi Iwamura, Naoto Kawamura","doi":"10.1145/3442536.3442541","DOIUrl":"https://doi.org/10.1145/3442536.3442541","url":null,"abstract":"Invisible digital watermarking, a technology for embedding information in digital content, is mainly used for copyright protection. In this paper, we proposed a method to identify images with invisible discrete cosine transform (DCT)-based watermarking and its vulnerability. Prior to watermarking, the images were normalized to 256 × 256 grayscale, and the network was generated through transfer learning; ResNet-18 was used to classify the input images as “watermarked” or “unwatermarked.” According to our results, the accuracy of the network, when classifying images into both classes, was as high as 99.80%. Furthermore, the testing accuracy of a network designed to detect the embedding location of the watermark in the frequency domain was 97.97%. It should be noted that the networks were ineffective when the DCT block size of the input images differed from that of the images in the training set.","PeriodicalId":353230,"journal":{"name":"Proceedings of the 2020 3rd Artificial Intelligence and Cloud Computing Conference","volume":"87 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124979229","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Indoor Location Estimation Based on Inverse Fingerprints at Multiple Points in Time and Moving Distance","authors":"Yuto Oue, T. Umezawa, Noritaka Osawa","doi":"10.1145/3442536.3442548","DOIUrl":"https://doi.org/10.1145/3442536.3442548","url":null,"abstract":"In indoor position estimation, it is generally difficult to estimate the coordinates accurately using indoor localization methods such as fingerprinting, triangulation and pedestrian dead reckoning alone. This is because each method has its own advantages and disadvantages due to differences in characteristics. Therefore, researches are being conducted to combine the methods to compensate for the disadvantages of each method and to estimate the position with a high accuracy. This paper proposes an indoor localization method that uses consolidated fingerprinting, which combines inverse fingerprinting and dead reckoning differently from conventional methods using combinations of them. The proposed method is based on inverse fingerprints at multiple time points and displacements or distances of movements between the time points. An experimental evaluation shows that the root mean square error of estimated positions is 0.42m for the proposed method and 1.09m for the inverse fingerprinting. The use of the proposed method reduces the error by about 61%.","PeriodicalId":353230,"journal":{"name":"Proceedings of the 2020 3rd Artificial Intelligence and Cloud Computing Conference","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117101710","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}