Main Project Office

蔡宗翰2.jpg

Professor

Tsai Tsung-Han

 

PI of MOE AI competition and labeled data acquisition project

Department of Computer Science and Information Engineering

at National Central University

Director of Taiwanese Association for Artificial Intelligence

2018生活照 吳世弘2.jpg

Assistant Professor

Wu Shih-Hung

co -PI of MOE AI competition and labeled data acquisition project

Department of Computer Science and Information Engineering

at  Chaoyang University of Technology

李宏毅2.jpg

Associate Professor

Li Hung-i

co - PI of MOE AI competition and labeled data acquisition project

 

 

Department of Electrical Engineering at  National Taiwan University

​Hsu Chia-Ling

Project Assistant

Phone : (03)422-7151 #35323

E-mail: jialing210@g.ncu.edu.tw

肺腺癌病理切片影像之腫瘤氣道擴散偵測

鍾翊方.jpg

Associate Professor

Li Chi-Chun

principal investigator

Department of Electrical Engineering

at National Tsing Hua University

葉奕成.jpg

Assistant professor 

Hsu Chih-Chung

co-principal investigator

Department of Information management

at National Pingtung University of Science and Technology 

應用人工智慧分析數位病理影像是近年來新興且極具潛力的領域。透過分析腫瘤病理影像,可以取得豐富的腫瘤相關資訊,應用於臨床醫療。 「腫瘤氣道擴散(spread through air spaces, STAS)」是近年新發現肺腺癌的病理特徵,指的是腫瘤細胞從腫瘤的邊緣沿著肺泡空腔,往鄰近的正常肺臟組織擴散。腫瘤氣道擴散對肺腺癌病患接受手術後復發的風險有重大影響,已是目前肺腺癌病理檢查的例行項目。 針對肺腺癌H&E染色數位病理全切片影像,本競賽提供在腫瘤外的感興趣區域(region of interest, ROI)以方框及不規則形狀像素層級之STAS標註資訊,進而舉辦兩種競賽:(1) 運用物體偵測作法於找尋STAS;(2) 運用影像分割作法於切割STAS輪廓。

本競賽提供豐富獎項、雲端運算資源、軟體工具和訓練課程,協助參賽者在競賽過程中,具備堅強的實力及工具取得佳績,像是有機會運用全台最大的 AI 高速 GPU 運算雲端資源-TWCC,協助跨節點高效能運算快速執行大規模的應用程式,透過AIHPC高效能加速縮短計算時間、優化模型以產出最精準的競賽結果。

農地作物現況調查

王驥魁.jpg

Associate Professor

Li Chi-Chun

principal investigator

Department of Electrical Engineering

at National Tsing Hua University

林昭宏.jpg

Assistant professor 

Hsu Chih-Chung

co-principal investigator

Department of Information management

at National Pingtung University of Science and Technology 

呂學展.jpg

Assistant professor 

Hsu Chih-Chung

co-principal investigator

Department of Information management

at National Pingtung University of Science and Technology 

農地作物現況調查可使用人員搭配相機於現地拍照紀錄,然而農地區域廣泛、我國之坵塊分割細碎且分佈零碎,所獲取之照片影像資料龐大,轉換為可用於管理操作之資訊,極度耗費人力、時間。AI 技術對於影像判識工作近年已有長足之進步,適合導入農地作物現況調查之工作程序,加速農政單位獲取相關資訊。
目前雖然有完整的民生/工業/醫療等AI資料集,但在農業數據相對缺乏,故在未來AI智慧農業需求上,將需要投入大量的專業人力進行農業數據蒐集及分析作業;透過本競賽,將協助學生瞭解農業資料集及農產業影像辨識的應用需求,並培育學生應用AI進行農業領域影像辨識的經驗及技術能力。

Automatic Positioning and Application of Rice UAV Full Color Image Plant Position

黃仁竑.PNG

Associate Professor

Li Chi-Chun

principal investigator

Department of Electrical Engineering

at National Tsing Hua University

江振國.jpg

Assistant professor 

Hsu Chih-Chung

co-principal investigator

Department of Information management

at National Pingtung University of Science and Technology 

臺灣蘭花栽培歷史悠久、品種繁多,產量及品質在國際上受到肯定,臺灣有著領先全球的蘭花育種研發,擁有世界上最多的蝴蝶蘭品種,九成的蝴蝶蘭用於出口,使其成為國內精緻農業金額首位。但因農業生物科技的進步,大量組織苗的繁殖,影響新品種的研發,加上其它國家積極投入育種生產,育種廠商大都有其專注培養的品種,由於各蘭花品種的型態較為類似,往往需要專業人士才能進行區別,目前全球均未有蝴蝶蘭品種的辨識軟體與技術,本競賽欲透過舉辦蘭花種類辨識及分類競賽,邀集具AI、機器學習、深度學習專長的專家與高手,訓練出高辨識率的蘭花品項影像辨識模型,除了能提升產業競爭力外,更讓社會大眾對於蘭花這類高經濟作物有更多的認識,進一步提升蝴蝶蘭的銷售與產值。

無人機飛行載具之智慧計數(農作物、車輛與人群計數)

謝君偉.jpg

Associate Professor

Li Chi-Chun

principal investigator

Department of Electrical Engineering

at National Tsing Hua University

莊仁輝.png

Assistant professor 

Hsu Chih-Chung

co-principal investigator

Department of Information management

at National Pingtung University of Science and Technology 

張明清老師.jpg

Assistant professor 

Hsu Chih-Chung

co-principal investigator

Department of Information management

at National Pingtung University of Science and Technology 

近年來科技高度發展所帶來之影響,已逐漸深入人類生活環境。無人飛行載具提供更寬廣的視野及更高度的移動性和靈活性,至今已被應用在許多不同產業領域,如地理資訊蒐集、交通監控、物品運送、通訊網路中繼站等不同類型。本次計畫預計將由國立陽明交通大學智慧科學暨綠能學院謝君偉教授、資訊學院莊仁輝教授及美國紐約州立大學奧爾巴尼分校張明清助理教授合作,著重於 AI 與影像辨識及無人飛行載具應用,期望參賽者以日常生活可能遭遇之問題為出發點,將深度學習原理作為基礎,運用相關人工智慧核心知識深入發展應用,進而將無人飛行載具相關技術應用於實際環境中,結合不同領域之人才並發揮創意,將平時所學專業知識確實落地,提升個人及
團隊之實作能力與競爭力。
本次競賽主題為 AI 與影像辨識-無人機飛行載具之智慧計數(農作物、車輛與人群計數),無人機載具有高度移動性以及遠距遙控功能,能夠快速且輕易到達不容易接近的區域,搭配高解析度相機即如同鷹眼般,能從空中俯視地表,並將地表一切變化詳實記錄在影像中而不遺漏,這種高機動性與廣泛應用,讓世界各國皆視無人機為下一代工業 4.0 革命中不可或缺的重要部分。目前國內尚無此空拍影像分析之比賽,此計畫將以無人機空拍影像為基礎,運用深度學習原理等相關訓練模組,進行高空農作物、車輛與人群數量技術與辨識。藉由此比賽,利用做中學的方式,進而掌握關鍵空拍分析技術,並培養前瞻 AI 研發人才,導引高階研發人才至各社會階層進行貢獻與服務,為國家高科技產業帶入新動力。

自然語言理解的解釋性資訊標記計畫

黃瀚萱.png

Associate Professor

Li Chi-Chun

principal investigator

Department of Electrical Engineering

at National Tsing Hua University

機器學習模型的可解釋性(explainability)是人工智慧技術在落實與應用時,備受期待的項目。如果模型能夠在精準預測之餘,同時提供佐證其預測行為之依據,人類將有機會事先發現模型判斷的錯誤,大幅提高人類對模型之信賴,使得人工智慧技術更能落實到關鍵性的決策場域。
目前的自然語言處理模型,對於這類的任務已能達到頗高的效能,顯示近期的深度學習技術已能掌握部份語言推理能力。然而,模型在提供準確的推理(即三分類任務)之餘,究竟是如何得到該推理之結果,其中的解釋性要素則仍然未有充份的研究。
本計畫以自然語言推理為目標,希望能讓模型在預測文句之間蘊含、矛盾、無關等邏輯關係的同時,找出文句之中關鍵性的片段,作為預測結果的佐證資訊。這樣的資訊一方面可以提供機會讓研究人員更了解模型內部的行為,促進自然語言處理的研究。同時也可在將來應用人工智慧系統時,提供模型的判斷依據,供人類評估模型該次判斷的可靠程度。

繁體中文場景文字辨識競賽

su-bo-ci.jpg

Associate Professor

Li Chi-Chun

principal investigator

Department of Electrical Engineering

at National Tsing Hua University

張明清老師.jpg

Assistant professor 

Hsu Chih-Chung

co-principal investigator

Department of Information management

at National Pingtung University of Science and Technology 

Recently, with the rise of artificial intelligence, the introduction of precise technology into the concept of smart value-added technology in agriculture will not only help increase productivity, but also improve population shortages and income from agricultural production. It is expected that high automation and standardization will be achieved through AI technology, which improve output value and effectively carry out industrial transformation.

 

This goal hopes to introduce concepts based on automated and accurate AI image recognition technology. In the future, we will use Taiwan's local database to build an automatic mango sieve fruit system. In the future, we hope to cooperate with government units such as the Agricultural Committee to complete the mango intelligence resume. This provides a simple and clear mango purchase mechanism for the general public in Taiwan, while at the same time pulling the local Aiwen mango brand refinement.

 

Through the competition, you can quickly grow this related technology and train related intelligent agricultural talents. By grasping the three elements of technology cultivation, technology traceability and brand value, in order to improve the quality of local mangoes and establish the brand value of Taiwan mangoes. it will not only help Taiwan's overall outward expansion of competitiveness, but also have the opportunity to establish with other countries. The connection has raised the popularity of Taiwan's high-economic vegetable and fruit brands and even the internationalization of planting indicators. At the same time, in response to this theme that is closely related to life, Taiwan's AI talents have also been cultivated, which has spread to different agricultural crops and laid the foundation for smart agriculture.

Automatic Positioning and Application of Rice UAV Full Color Image Plant Position

劉力瑜.jfif

Professor

Liu Li-Yu

principal investigator

Department of agronomy

at

National Taiwan University

黃乾綱.jpg

Associate Professor

Ku Lun-Wei 

co-principal investigator

Department of Engineering Science and Ocean Engineering

at

National Taiwan University

With the rapid development of software and hardware, the application of artificial intelligence has been successfully applied in a wide range of fields. However, the application of artificial intelligence in agriculture is quite limited. In addition to the long growth cycle of crops and the difficulty of accumulating large amounts of data, the data labeling highly relies on agricultural experts, which makes it seriously insufficient for training data. Aerial telemetry images and Internet text are a large amount of information that is relatively easy to obtain.

In response to the development of "new agriculture" from the government, this project first predicts the use of high-resolution drone full-color images for plant location labeling, and the release of plant epidemics from the open platform of Council of Agriculture Executive Yuan R.O.C. In the future, based on the labeling results, artificial intelligence can be used to build rice field plant identification and text identification analysis modules to enhance the soft power of Taiwan’s future agricultural development.

The results of the labeling can be applied to the application of drones to spray plant protection agents in small quantities in the future to achieve the dual purpose of saving costs and maintaining environmental sustainability. It is also in line with the recent goals of the application of unmanned plant protection machines actively promoted by the Council of Agriculture Executive Yuan R.O.C.
 

Decision Analysis of Medical Information and Construction System of Dialogue Corpus

高宏宇.jpg

 Professor

Kao Hung - Yu

principal investigator

Department of Computer Science and Information Engineering

at National Cheng Kung  University

柯乃熒.jpg

Professor

Ko Nai-Ying

co-principal investigator

Department of Nursing

at  National Cheng Kung  University

According to the provisions of the Health Insurance Portability and Accountability Act, in the clinical medical text records, the content of the patient's private information must be erased or modified. In the outpatient medical-patient dialogue data, there are many private content of the people seeking medical treatment. Such a large amount of data requires an automated way to identify these private content, which facilitates the work of medical staff and accelerates the establishment of medical big data.

This competition provides outpatient dialogues and consultation dialogues of related interviews collected from the outpatient clinics of Chengda Hospital. The private content and types of the dialogues are manually marked. And the data is divided into training set, construction set (development set) and test set.

The main goal of this competition is to identify and extract content containing private information from the dialogue between doctors and the public, and to classify the privacy type of the content.

Taiwan Fruit Image Recognition Competition

-Take Aven Mango as an example

李祈均.jpeg

Associate Professor

Li Chi-Chun

principal investigator

Department of Electrical Engineering

at National Tsing Hua University

許志仲.jpg

Assistant professor 

Hsu Chih-Chung

co-principal investigator

Department of Information management

at National Pingtung University of Science and Technology 

Recently, with the rise of artificial intelligence, the introduction of precise technology into the concept of smart value-added technology in agriculture will not only help increase productivity, but also improve population shortages and income from agricultural production. It is expected that high automation and standardization will be achieved through AI technology, which improve output value and effectively carry out industrial transformation.

 

This goal hopes to introduce concepts based on automated and accurate AI image recognition technology. In the future, we will use Taiwan's local database to build an automatic mango sieve fruit system. In the future, we hope to cooperate with government units such as the Agricultural Committee to complete the mango intelligence resume. This provides a simple and clear mango purchase mechanism for the general public in Taiwan, while at the same time pulling the local Aiwen mango brand refinement.

 

Through the competition, you can quickly grow this related technology and train related intelligent agricultural talents. By grasping the three elements of technology cultivation, technology traceability and brand value, in order to improve the quality of local mangoes and establish the brand value of Taiwan mangoes. it will not only help Taiwan's overall outward expansion of competitiveness, but also have the opportunity to establish with other countries. The connection has raised the popularity of Taiwan's high-economic vegetable and fruit brands and even the internationalization of planting indicators. At the same time, in response to this theme that is closely related to life, Taiwan's AI talents have also been cultivated, which has spread to different agricultural crops and laid the foundation for smart agriculture.

Singing and Chord Recognition Competition

張智星.jpg

Professor

Chang Chih-Hsing

principal investigator

Department of Computer Science and Information Engineering

at National Taiwan University

With the development of the times, the way of listening to music has shifted from the previous CD to various online music platforms.

The 2016 IFPI report pointed out that the output value of digital music has officially exceeded the output value of physical music, and the output value of physical music is decreasing year after year, showing that the trend is on the digital music side.

The development of digital music has driven many related AI intelligent applications, including original song identification, humming song selection, and song classification. Among them, Line Music, KKBOX, and Spotify have successively established machine learning or artificial intelligence departments, which analyze the user's music preferences through song content and user's listening habits. The various AI services above are provided to allow users to easily listen to their favorite music, thereby generating a space of added value.

Through the development of digital music, various music-related services such as song recommendation and humming search have also become more and more. Domestic companies such as KKBOX and foreign countries such as Spotify have also set up their own machine learning departments to do research and expansion of automatic song recommendation.

Such services can be divided into two parts in the study of machine learning. One is to make recommendations based on user behavior, and the other is to make recommendations based on the essence of the song. In the essence of the song, many basic feature method studies are necessary. The basic elements of the song's main melody, chords, song structure, genre and tempo, constituting how a song should be classified and recommended. So the talents who familiar with these basic elements of music analysis and machine learning methods in the current music industry are urgently needed to be cultivated.

Competition on Machine Reading of

Artificial Intelligence Papers

陳銘憲.jpg

Distinguished Professor

Chen Ming-Hsien

principal investigator

Department of Electrical Engineering at  National Taiwan University

林守德.jpg

Professor

Lin Shou-Te

co-principal investigator

Department of Computer Science and Information Engineering

at National Taiwan University

葉彌妍副研究員.png

Associate Research Fellow

Yeh Mi-Yen

co-principal investigator

Institute of Information Science, Academia Sinica

In this competition, we will let students try to solve a problem that has long been a headache for computer science researchers with the technique of semantic analysis: "How to design a system that can automatically read the abstract of the paper, mark and unify the invention, use or application To compare algorithmic systems. "

It will be an unavoidable attempt to let the machine automatically sort out these algorithms that are constantly being developed. Even if the manpower is affordable, letting the machine automatically integrate related algorithms will allow researchers to free up time to do more meaningful things. 

Competition on News Stance Retrieval

News on controversial issues has always been the focus of viewers' attention and discussion. News media often need to report different positions. If a large number of news documents can be quickly searched for news with a specific position on various controversial issues, It will not only help people understand the different perceptions and different values ​​of these issues from different positions, but also be of great reference value to the decision-making process.

This competition encourages students to use creativity and technology to solve the challenges of data retrieval and opinion exploration applications. It also looks forward to training and cultivating talents in artificial intelligence fields such as information retrieval, natural language processing and machine learning in China through the competition.

Associate Professor

Cheng Pu-Jen 

principal investigator

Department of Computer Science and Information Engineering

at National Taiwan University

古倫維.jpg

Associate Research Fellow

Ku Lun-Wei 

co-principal investigator

Institute of Information Science, Academia Sinica

Competition on Biomedical Paper Analysis

Distinguished Research Fellow

Hsu Wen-Lien 

principal investigator

Institute of Information Science, Academia Sinica

This competition will use natural language processing technology as the core, and the open competition corpus will enable students to apply artificial intelligence technology to basic clinical medical history data analysis to advanced bioinformatics document corpus.

Students participating in this competition will have a practical understanding of the processes and techniques of analyzing biomedical data and reflect the translation of medical thinking, so that students can understand the big thinking of connecting basic medicine, bioinformatics research, and clinical treatment with natural language processing technology.