This research project aims to establish an intelligent agricultural cultivation support system, by integrating unmanned aerial vehicle (UAV) surveillanceartificial intelligent (AI) analytical techniques.

With the rapid advancement of artificial intelligenceIoT, many solutions have been successfully implemented. However, for very large biomedical image computations, such as MRI’s, the deep learning/ training will be a lot more time consuming.

With our autonomous UAV systems, we can effectively reduce many high-risk and high-cost inspections which require high manpower, such as inspections of dams, river, coastline patrols, etc.. Operators only need to setup and launch the system to allow the UAV to complete the designated tasks independently.

AI-enabled Service Assurance Platform for 5G Vertical Application: The platform can help customers quickly import 5G vertical applications by providing overall customized solutions, including equipment evaluation, network deployment, application importing, network maintenance, and operation optimization. The platform can bridge the gap between the telecom industries and the vertical application industries and provide a total solution for the industries to import 5G vertical applications.

AI Deep Compression Toolchain: To enhance the computing power and lower the power consumption of edge devices, this toolchain delivers a 120x AI model compression ratio from original floating-point inference models, compress into fixed-point and bit-accurate edge models with a smaller than 1% accuracy loss. We also provide an open-source image bank and fully automatic labeling tool, available for free application and use at https://www.aicreda.com/.

Tool wear and health condition monitoring during the processing: The tool wear monitoring technology developed by our researching team is specifically designed to analyze whether the tool is broken, collapsed, etc., and to estimate the remaining useful life(RUL) of the tool according to the working conditions of mass processing. By acquiring the vibration signal data with three-axis accelerometers installed on the machine tool, this technology could determine whether the current tool cutting vibration has exceeded the safety range by plotting a control chart. Once it exceeds the safe range, the current tool processing state will be assumed as abnormal. It gives users a reference to replace the broken tools immediately to prevent continuing processing, which causes vast loss such as poor quality of workpieces. In addition, this technology allows users to build models for distinct working conditions to predict the RUL of tools. It could allow users to evaluate the current health condition of tools and schedule the time to change the tool.

We have developed a movement assistive system using EEG-controlled functional electric stimulation system. Users’ movement intentions are recognized by brain computer interface, implemented by deep learning network, and the detected users’ intentions are then translated into commands to trigger a functional electric stimulation (FES) device to activate users’ particular movements, such as grasping, hand raising, holding a cup, etc. In our study, we have designed our own wireless dry-electrode EEG system and our own FES system.

div style="text-align: justify"Quantum key distribution exploits the transmissiondetection of single photons’ quantum states to generatedistribute secure keys, allowing absolutely secure communication. We use a self-developed miniature single-photon source, together with the single-photon wavepacket engineering, to implement a highly efficient protocol of differential-phase-shift quantum key distribution. Using the Campus Fiber Network between National Tsing Hua UniversityNational Chiao Tung University, we demonstrate Taiwan’s first outdoor quantum key distribution. The technology not only benefits the development of distance unlimited absolutely secure communication networks for the commercialmilitary uses, but also opens up new opportunities for the R&Dmarkets in industry./div

div style="text-align: justify"Constructing a functional connectomeits computational model is a crucial step toward understanding the mechanisms of brain functions. To achieve this goal, we developed two correlated technologies: (1) An all- optical physiology (AOP) that is capable of millisecond volumetric imagingaccurate stimulation in living animal brains. This system allows us to establish functional connectomeneural coding with a single-cell resolution. (2) A cellular-level spiking neural circuit simulation system that is capable of tuning itself based on the input data from the AOP system. We have demonstrated our technologies in the Drosophila late visual systemwill apply them in the brains of larger species such as mice. We expect that our technologies will be able to greatly enhance our knowledge of the brain operation principles. Our 3D all-optical physiology (AOP) platform incorporates single-photon point stimulationtwo-photon high-speed volumetric recordings (Optics Letters 2019, "Editors pick"). We have demonstrated its effectiveness in studying the anterior visual pathway of fruit flies (iScience2019). In comparison, contemporary high-speed AOP platforms are limited to single-depthdiscrete multi-plane recordings that are not suitable for studying functional connections. Our high-resolution computational model is constructed based on the combination of static connectomeAOP data,is much more realistic than the existing models. Our work aids establishing in-vivo 3D functional connectomescomputational models of the brains, thus provides insight into the mechanisms of brain functions./div

Artificial Intelligence Techniques Aquaculture Management System: The system uses the omni-IoT system to collect big data, provides AI algorithm for each module, included AI ​​feeding module, fish body length weight measurement module, smart submersible cage module, and provides better fish growth control , to reduce residual bait, to improve survival rate of fish, to save manpower, to reduce the threshold, cost and risk of smart cage culture operation. The AIoT system of our team is mainly self-made, which greatly saves costs and is modularized. Aquaculturist can choose modules to use allow young fishermen to profit easily even if they do not have a lot of farming experience.

Recently, Deep Reinforcement Learning (DRL) has been applied to many AI applications. One of the successful achievements is the AlphaZero, called the Zero method in this project, was presented to learn without human knowledgesurprisingly surpass all the human playersall the AI programs.

Textual Advertisement Generator: Given any limited specifics of any product, AI Advertisement Producer can automatically generate tons of top-quality descriptions and advertisements for the product in just one second. And not just one copy is produced. With deep learning and natural language processing technologies learned from millions of existing samples, our AI model can produce various styles of advertisements at the same time for users to select. It will be a big helper or a virtual brainstorming partner for any brands or advertisers to create their advertisements.

For video anomaly detection, we apply pretrained models to obtain the foreground and the optical flow as ground truth. Then our model estimates the information by taking only a single frame as input. For human behaviors, we take the human poses as input and use a GCN-based model to predict the future poses. Both the anomaly scores of these two works are given by the error of the estimation. For defect detection, our model takes patches of the image as input and learns to extract features. The anomaly score of each patch is given by the distance between the patch and all the training patches.

AI農情調查之UAV群眾協作平台後台支援自動鑲嵌建模，更具備四項突破技術：(1)巨量影像格化技術；(2)平行運算技術；(3)任務規格標準化；(4)UAV任務媒合。致力於打造空中UBER協作服務，未來能應用於農作物分佈調查、大範圍災情調查、農業保險、農地違法使用調查與休耕補助調查等面向。

Development of AI Platform for Smart Drone - Intelligent Flight: Due to its high mobility and the ability to fly in the sky, the drone has inspired more and more innovative applications/services in recent years. The goal of this project is to resolve the problem of blindly flying an unmanned aerial vehicle (UAV, which a drone in our case) when it is out of human sight or the range of wireless communication, and three major research and development directions will be considered in this project. Three artificial intelligence (AI) technologies, namely, smart sensing, smart control, and smart simulation, are applied in this project. Smart sensing - a flight system is developed, which can avoid the obstacles, complete a flight mission, and land safely. Smart control - an intelligence flight control system and a light-weighted somatosensory vest are developed. Smart simulation - a cost-effective training system and a 3D model simplification method are designed.

Taiwan was formerly the Shrimp Capital of the World. With the changing of the timesincreasingly updated technology, AI smart farming is the new milestone for Taiwan’s agriculturefisheries.

div style="text-align: justify"Taiwan was once the shrimp capital of the world. Though just a small island, the technology, abundance,reputation was globally known. Over time, the changes in water qualitythe environment led to lessless shrimp. AI may be the innovative solution needed to revive the industry./div

Through the technique, Sampling and Prediction of Lithography Overlay Errors, the cost and time of overlay error measurement can be reduced to improve the process efficiency. We identify key sampling through clustering and machine learning models, and design a new sampling algorithm in photolithography process. Due to the complexity of wafer and many training factors of wafer data, we combine the clustering algorithm with incremental learning to meet customers＇ unique needs and achieve the goal of optimally sampling and reducing the costs.

Our system provides system identification, servo tuning, and feed axis diagnosis for machine tools. Also, executing remote function verification, and data collection are included. All these functions were actually tested verified by the co-operation manufacturer. It help operators or engineers to maintain and adjust the machine so that the feed drive system match. During this critical time of the COVID-19 epidemic, the proposed system provides a zero-touch remote diagnostic function and method for engineers to maintain the machines in the processing plant can effectively reduce costs and time.

1.2 kV SiC Super Junction MOSFET: We have developed the first 1.2 kV SJ MOSFET in Taiwan in 2020. Its threshold voltage is 3.5 V, saturation current is about 12 A, and the specific on resistance is 8mΩ*cm2.

Web interface of CPS platform: Although many factories have introduced MES and scheduling systems, they have not been integrated with the CPS architecture. The transmission of information is not transparent and automated, and it is difficult to ensure the relevance and continuity of the information, which affects the control of the entire production process and the on-site real-time information. Our integrated CPS structure enables the project to obtain the best dispatch of the machine and the shortest delivery time. We also use DDS to receives real-time on-site information and presents it as a platform, which enables control the real-time on-site situation. We constructed an advanced plastic mold & molding service platform, combining CPS and cloud architecture, based on the production life cycle, integrating heterogeneous equipment and systems, and it achieves the goals of real-time monitoring of the manufacturing process, rapid iteration and dynamic optimization.

The DNN models of this technology include a LithoNet, an OPCNet, and a layout novelty detection network. LithoNet is a learning-based pre-simulation model for layout-to-SEM contour prediction, and OPCnet is a dual network of LithoNet for photomask optimization. Integrated with a well-trained LithoNet, our layout novelty detection network, consisting of a self-attention guided LithoNet and an autoencoder, can check if there are layout patterns easily resulting in local distortions in contours of metal lines based on multi-modal (global-local) feature fusion.

This technology aims to automatically extract and define features from a data-driven framework of data engineering, machine learning, and ensemble learning algorithms to analyzing the big data collected from equipment sensors and quality characteristic measurement results. This technology provides prediction results of unmeasured products quality characteristics and its confident score base on the extracted feature for advanced process control.

Reliability of Cu-to-Cu direct bonging: In order to realize the reliability of Cu-to-Cu direct bonding interconnects, we adopted nanotwinned Cu (nt-Cu) electroplating process to fabricate nt-Cu microbump for Cu-to-Cu direct bonding. Figure (a) is the photo of chip after bonding. Underfill (UF) dispensing was used to protect Cu microbumps. The cross-section SEM image of Cu microbump is shown in figure (b). Some chips underwent the reliability tests, including temperature cycling test (TCT) and electromigration (EM) test. To further study the failure mechanism, finite element analysis (FEA) was done to know the stress and current distribution during reliability test. This research could provide fundamental understanding for Cu-Cu joints with organic dielectric hybrid bonding.

div style="text-align: justify"According to the statistics of World Health Organization (WHO) in 2018, Chronic Obstruction Pulmonary Disease (COPD) has killed 3 million peoplebecome the 3rd place among the top 10 leading causes of death in the world, which means one person dies for this fatal disease every 10 seconds. In Taiwan,chronic lower respiratory disease is at the 7th place among the top 10 leading cause of death. There are over 5000 people died for obstructive pulmonary every year.

Use NLP techniques to realize the automatic coding of ICD10. According to the input of the patient’s age, gender, medical order, admissions, progress note, surgical records, discharge, ICD-10 diagnostic code and ICD-10 disposal code, perform machine learning model training and code prediction.

Intelligent Portable Neurofeedback Training System: Highlights of our product: 1. High accurate EEG value (with rigorous hardware and software design and verification) 2. real-time feedback by computer screen or APP 3. can be used at home (easy to ware, easy to use)

Combined with millimeter wave radar detection of chest undulation breathing mode and heart rate, continuous blood oxygen detection, active disease record of chat robot, and mobile phone analysis of 30 second sitting and standing alternate activity frequency mode, a set of personalized respiratory capacity benchmark is established through AI modeling, which can be applied to zero-contact respiratory physiological monitoring and useful for infectious disease ward, epidemic prevention hotels, centralized quarantine centers.

The dialogue system is the main subsystem of the visually impaired navigation system, which provides destinations for the navigation system through multiple dialogues. We use the knowledge graph as the basis for reasoning. In terms of close-range navigation, deep learning technology is used to develop RGB camera detection depth algorithm, indoor semantic cutting algorithm, integrated detection depth estimation and indoor semantic cutting in indoor obstacle avoidance, etc. The whole system uses the CellS software design framework to integrate distributed AIoT systems.

The IHC staining intensity grading algorithm combines a series of complementary processes of deep learning and image processing in order to overcome the difficulties in cell boundary segmentation and grade definition, and provides a cell-based IHC staining intensity grading and quantification capability. The proposed method serves as a useful assistive tool for physicians in performing accurate staining quantification in tissue microscopic images.