Artificial Intelligence

Combing the Government Open Datasets to Build the Location Selection Decision Support System—The Local Consumption Ability and Neighboring Industries Perspectives [ 下載 PDF ]

Yi-Ning Tu, Wei-tse Hsu, Hsiang-Chin Huang, Ming-Kai Hsu, Yu-Hsiang Lin, Jian-Jie Hong

This study hope to combine the open data and current machine learning methods to establish “location selection decision support system”, which can provide suggestions to both landlords of the store, and the coming shopkeepers. For collecting store information, this work connects 3 different sources of open datasets. Furthermore, to quantify the surrounding information of a store, this study measures the distance between locations predicted to landmark or to each type of stores. For discriminating whether or not a type of store congregates, this work adopts Moran's I spatial autocorrelation analysis. spatial autocorrelation analysis. This study utilizes the Random Forest Importance to identify the key factors of 30 distinctive types of store, and apply k-nearest neighbour for the foundation of recommendation. As the results, this work shows in Precision, the proposed method is at least 26.8% higher than other classification algorithms.

Performance Measurement and Analysis Techniques for High-Performance Distributed Deep Learning Systems [ 下載 PDF ]

Cheng-Yueh Liu, Shih-Hao Hung

Deep learning has been widely used to develop artificial intelligence applications recently. Academia and companies race to find new deep learning algorithms and solve important scientific/engineering problems, as well as develop applications for business and daily use with high-performance heterogeneous computing systems. Hence, the system infrastructure to accelerate the development and deployment of deep learning applications using a largescale computing cluster with state-of-the-art accelerators is not only critical, but also contains many complexes, challenging research opportunities. In this article, we introduce our research work on performance measurement and analysis techniques, which provide essential information to help construct the aforementioned system infrastructure and enable the system designer to examine the factors that may affect the performance of various neural network models on mainstream deep learning frameworks. We show how our tools can be used to investigate on performance bottlenecks and optimize the system performance by adjusting the the system architecture, the parameter update mechanism, and the compiler options.

The Development of Vision-based Pedestrian Detection and Tracking Technology [ 下載 PDF ]

Chi-En Hu, Jonathan Guo, Yuan-Kai Wang

As of today, Artificial Intelligence has gained a lot of attention and is currently being developed rapidly. As one of the key technologies, Computer Vision has become the focal point of many researches and applications, due to the rapid development of pedestrian detection and tracking methods used in Intelligent Surveillance Systems and Self-driving Cars. Pedestrian detection and tracking are two different technologies, yet inseparable when used in applications. Due to the rise of deep learning, pedestrian detection has made great progress in accuracy, which also has driven the improvement of pedestrian tracking. In this paper, we in detail go through the introduction of pedestrian detection and tracking technologies, including the development of both technologies, and also the introduction of online multiple object tracking systems.

Machine Vision and Deep Learning Based Defect Inspection System for Cylindrical Metallic Surface [ 下載 PDF ]

Eugene Su, Yuan-Wei You, Chao-Ching Ho

Since the Alex Krizhevsky team won the championship of ImageNet Large Scale Visual Recognition Competition in 2012, deep learning has set off a new wave of artificial intelligence field, with all industries and research groups competing for deep learning field. However, the open question is whether the deep learning is the magic solution in a fairy tale or not. In this paper, we discuss how to employ the deep learning in the field of automated optical inspection industry, and analyze the skills of directly applying the current deep learning model to the defect detection based on the optical inspection method. Compared to the deep learning in the fields of speech recognition, natural language processing, medical or other computer vision applications, few literatures are found in discussing the defect detection based on deep learning. Therefore, this paper offered the solution of applying the deep learning in the field of defect detection for golf clubs based on the optical inspection method.

"Disaster Response: Artificial Intelligence in Swarm Ground, Surface, Aerial and Underwater Robot" [ 下載 PDF ]

Min-Fan Ricky Lee, Li-Jung Yang, Wei-Yi Kong

After the Fukushima nuclear disaster and the Wenchuan earthquake, the relevant government agencies in various countries realized the urgency of disaster response robots. There are many natural or man-made disasters in Taiwan, and it is usually impossible to dispatch relevant personnel to search and rescue or exploration at the earliest time. This paper proposes that ground, marine, aerial, and underwater robots collaboration in disaster response. Heterogeneous robots collect environmental big data from the scene and then use artificial intelligence analysis in the cloud to make the best decision. Related tasks include target monitoring, search and rescue. This paper aims at the safety, security and rescue in disaster response, and thus responding to various actual disasters; reducing the casualties, addressing the shortage of manpower, and large-scale search and rescue.

Introduction to Privacy-preserving Mechanisms for Artificial Intelligent Systems [ 下載 PDF ]

Peter Shaojui Wang

With the rise of artificial intelligence, many intelligent services have benefited our life; one of them is Big Data technology, which has helped the industry and the academia get the trends that are not easy to be observed in the past. However, it also has raised the serious privacy concerns while the massive amounts of our personal data are routinely collected in this Big Data era. This paper introduces several privacy-preserving mechanisms for artificial intelligent systems, especially Big Data systems, gives a review, and discusses all possible scenarios of them.

---

Fabrication of Micro Polarization Optics Device by Laser Bleaching and Study of Its Characterization by Microspectrophotometer [ 下載 PDF ]

Chun-Jen Weng, Chen-Liang Fan, Yu-Hsuan Lin, Yu-Hsin Lin

A grey scale laser bleaching method has been proposed in this study. A polarimetric microspectrophotometer (MSP) was built in this project in the first. Then, the two dimensional micro polarization optics devices with controllable extinction ratio was made from the polarizer film through laser bleaching via monitoring by this MSP. The devices made from conventional polarizer films have the special characteristics such as low cost, broadband, high dynamic range of extinction ratio and high acceptance angle. This work about setup of the MSP and single point laser bleaching scheme was done in the first. Optimized laser bleaching wavelength and the relationship between extinction ratio of bleached polarizer film and laser power was studied by the spectrometer. The polarimetric image distribution of bleached polarizer film was measured by the image camera. Next, two dimensional micro polarization optics devices based on linear polarizer film was fabricated with 2-axis linear stages controlled by LabVIEW program. A polarimetric MSP was used as real-time polarimetric analysis in order to control the extinction ratio and other polarimetric characters. These devices have the potential in the high grey scale mask, high density optical data storage, linear polarizer based grating and circular polarizer based grating.