TISF 2021

現今的大型活動，如：校慶活動、新北耶誕城等，缺乏互動性與參與感，其中原因大多是觀眾時常埋沒於手機中的社交軟體當中所導致。而我們的研究將利用此特性，探討大眾對於活動的觀點，搭配網路爬蟲抓取使用者的貼文，觀眾只需在Instagram、Twitter等社交軟體中發布文章，系統就會即時推播至活動中的大螢幕上，並且結合圖像辨識快速審核貼文，設計出一套能改善互動性低落的解決方案。研究中我們探討不同的網路爬蟲演算法、圖像辨識技術，及問卷調查…等來使作品更加精進，且搭配Line Bot、後臺管理，及常駐貼文…等功能來為各類大型活動量身打造，也能夠運用於政令宣導或文宣廣告等用途上，大幅提升活動的互動性與精采程度。

104年衛福部發布的研究分析中指出，台灣女性罹患乳癌機率過去30年大幅攀升。我們將面臨更多、更年輕乳癌患者。讓狀況更加嚴峻的是，患者在接受癌症治療時不但要忍受化療副作用，還有一定的機率喪失生育能力。因此，我們希望開發一個程式來協助研究人員及醫生測試乳癌治療藥物。藥物治療癌症的方式是透過與蛋白質進行連接，我們希望透過蒐集其SMILES化學式資料，轉換為MACCS形式，進而構成深度學習程式的訓練資料。因資料筆數較少，所以採用了ensemble learning 方法，製作多個weak model，然後再透過ensemble voting 將其合併並投票，產出準確預測結果。為了尋找positive data 與 negative data 之間比例，嘗試了四種不同比例，在分析結果後發現兩者比為1:20時，weak model的成效最好。我們在分析模型及測試結果使用了ROC precision 及 Spearman’s p進行分析，結果顯示模型有足夠的穩定性並能進行準確預測。

在資訊爆炸時代，效率閱讀、整理資料的能力越趨重要。身為高中生，學習時的閱讀量龐大，還須另外自己挑選重點句，重新整理筆記。因此我想如果可以讓電腦自動摘取文章的重點，就能幫助學生效率學習。 大多數現存的自動摘要研究適用於英文文本，本研究利用演算法抓取中文文章的摘要，使學生可以真正實用該演算法於日常學習當中。除此之外，此研究比較了不同方法摘要的準確率以及優缺點。

本研究以AI虛擬試衣系統(Virtual Try-on模型)為主題，透過深度學習技術，並結合幾何匹配模型，最後開發出試衣系統，將使用者上傳的照片，模擬成穿著新衣的模樣。首先，將人物原始圖片取出骨架節點，並生成人體遮罩以及保留人物頭部。而後將三個輸出合成為高維特徵圖。接著將目標替換衣物生成出依照人體姿態扭曲後的衣物圖片。最後於Virtual Try-on模型中，將人體高維特徵圖與扭曲衣物共同輸入，並經過運算後合成出穿著目標衣物之人體圖像。本研究結果發現，人物站姿單純，且雙手緊貼身側，以及拍攝角度為正面、衣服款式為短袖、背景色彩對比度較高與衣服圖案單純的原始圖片得出的結果圖片與測試圖片相似度較高。

圈圈叉叉可說是世界上最普遍流行的遊戲之一，而隨著不斷的改良，各式各樣的變化版層出不窮，但畢竟為民間自行改良的版本，由於缺少專人的檢驗或是電腦的推導，很多變化版都有明顯的先手優勢，導致遊戲的公平性大幅下降，如何最有效率的檢查各個版本的公平性便成為很重要的課題，在諸多版本中，有鑑於其龐大的遊戲樹(4.13*10^56)，我決定將終極井字棋作為本次的研究對象，本研究將結合決策樹以及極小化極大演算法訓練人工智慧，再由訓練結束的人工智慧彼此相互對抗多次，計算出遊戲的先手優勢，希望藉由本研究可以為終極井字棋提供一個客觀的先手優勢標準，並複製成功的經驗，運用人工智慧測試更多不同遊戲的公平性，另外，本研究也將探討此人工智慧在遊戲中對抗真實玩家的表現，並將結果用以佐證計算出的先手優勢的可信性。

第五代行動通訊（5th generation mobile networks）是現今科技發展的趨勢，新技術的出現也衍生出很多新的問題，在基地台點對點傳輸時，需要精確角度的天線才足以準確地接收高頻波短的毫米波，雖然現今已經有精密儀器能測量精確的方位角，但價格較高且使用方法複雜，面對數量龐大的5G基地台時，維修成本過高。本研究利用手機拍照得到天線與目標物相對角度，結合預先得知目標物的方位角，再經過數學運算即可得到精確的天線指向。本研究希望以隨手可得的手機，配合簡單的方法，可得到精確的天線指向，解決第五代行動通訊可能面臨的問題。

全球貧血人口普遍，然許多人並不了解自身是否罹患貧血；長期患有貧血的病人，亦需定期抽血檢驗追蹤是否有貧血惡化達到需接受輸血的程度。研究顯示，結膜之顏色與貧血有絕對關係，結膜越白則貧血越嚴重，醫師也常使用結膜顏色推測是否有貧血情形。若能設計手機軟體自動分割結膜影像並分析其顏色，將有機會推測受試者是否罹患貧血。本研究收集22位無貧血者及8位貧血病人，並獲得其近期血紅素數值。以手機取得受試者之眼睛影像後，成功設計程式以深度學習演算法完成結膜自動影像分割；並取結膜顏色計算其三原色比例，以圖表探討結膜顏色分布發現，貧血者和無貧血者之三原色比例呈現不同的分布。此結果代表AI技術將有機會透過結膜影像，判斷被拍攝者是否有貧血情形，未來若能增加研究人數並進行色調校正，將可設計網頁版或手機APP加以推測血紅素值。

情緒辨識是增進人際溝通的重要能力。如生命線、電話客服等應用情境缺乏表情、肢體語言等輔助時，單以語音進行情緒辨識有極高的實用價值。 本研究探討比較支持向量機（SVM）及卷積神經網路（CNN）兩種機器學習方法於訓練「AI語音情緒辨識」分類器模型的表現。我們採用SAVEE和RAVDESS兩個英文語音資料庫，並自行製作與標註「逼逼中文情緒語料庫」。研究結果顯示SVM對SAVEE資料庫單一情緒的辨識正確率達84~94%，個別錄音員正確率達75%，超越官網紀錄的73.7%。同時，實驗顯示深度學習的模型在訓練資料不足的狀況下，反而相對遜色。

本研究使用機器學習方法，改善年長者使用手機時觸控系統對於點按位置判斷之能力。首先設計實驗比較年長者使用手機時，點按位置及手勢判斷的準確率，接著收集年長使用者的觸控軌跡及裝置相關資料，並訓練模型以減少系統判斷的錯誤率和誤差幅度。再比較及分析不同機器學習模型對於本研究之資料的適用程度及經校準後點按位置準確率的提升，進而挑選出一個能夠最有效提升點按位置準確率的模型進行點按位置的預測。實驗過後選擇最有效提升準確率的Random Forest進行其他的校正實驗及分析。使用者點按位置的預測準確率能被有效提升，準確率能提高32.33%。而最終，將訓練後的模型套回實驗用的手機程式，系統判斷受測者的點按位置能從原本的63.98%提升至97.46%。

關於違規停車的研究,一直是重要的問題但因為牽涉層面太多太廣,實難以有效控制變 因的方式進行,所以一般多從「違規行為」的心理成因或者商圈地理位置,法規的制定著 手。但因為難以有效控制變因來進行研究。 因此本文旨在以資訊科學及數學的方法,建立一套分析的模式以進行研究。以連續小波 變換 CWT、Harr 離散小波變換 DWT、穩定離散小波變換 MODWT、差分整合移動平均自迴 歸模型 ARIMA 以及混合 MODWT+ARIMA 模型。對臺南市三個不同特性的行政里的共五個 聚類點,進行預測。原始的資料以週為單位,分成全日(allDay)及白天(Day, 08~16 時)兩種型 態,並以連續 46 週的資料預測接下來 4 週的結果。採用均方根離差 RMSE 相對平均值、標 準差、最大值的比值,進行預測品質的比較。

利用深度學習解決醫學問題一直是受矚目的研究主題。鑒於近期新冠肺炎疫情上升，有關新冠肺炎檢測的研究便成了熱門研究主題。目前，最有效的檢測方法是聚合酶連鎖反應 (PCR)然而，PCR耗時甚久且有人為誤差。因此，以X光影像圖透過深度學習來診斷並分級是一個有效率且安全的做法。在研究中，我們利用深度學習進行疾病診斷得到了極高的準確率(99.35%)、產生出疾病熱區及設計了新的分級系統( X-ray Severity Grading System , XSGS)，並將其用於嚴重程度分類，得到了可辨別的差異。

手語為聾啞人士日常溝通的工具，但對一般人來說這是一種難以理解的溝通方式。本實驗使用深度學習的 Yolov3 與 Yolov4 模型訓練37個國語注音符號手勢，然後再驗證模型對圖片、影片、即時(Real time)攝影辨識的正確率。 實驗結果顯示：Yolo v3 圖片辨識度效果還不錯，但影片辨識度很差，而Yolo v4 不管在靜態的圖片或動態影片都有不錯的辨識率，另外在即時的影像辨識也有不錯的效果。 雖然有部分符號的辨識度很低，但這可能是訓練時照片拍攝的問題，如果可以改進拍攝的數量和技巧，相信可以大幅提升判讀的準確率。

本研究主要考慮在盡量保留可讀性的情況下，找出將Imperative Programming程式碼對應的Functional Programming的程式碼並證明。 結果如下： 一、if statement 等價於由 ifte 函數所構成的程式碼，其中函數ifte定義在本文內 二、某些 for-loop statement 等價於由 foldl 函數所構成的程式碼 三、某些 for-loop statement 等價於由 map 函數所構成的程式碼

目前雖已有團隊針對口罩配戴辨識提出解決方案，但根據收集的資料，目前針對口罩辨識提出的方案多使用類神經網路yolo進行實作，使用於口罩辨識雖可達到有一定的效果，但對口鼻密合度不佳的細部狀態仍易有誤判的現象，以上狀況縱使民眾有配戴口罩，但若未與臉部、口鼻密合，仍有50％的空氣洩漏機會，無法有效阻隔飛沫傳染，形成防疫破口。而本研究在這樣的基礎架構下，再整合目前最強大的姿勢辨識之一的OpenPose，對這個題目進行更深一步地探討，以期達到更好的效果。本研究針對的改善的方向為當神經網路YOLO判定為有配戴正確的資料時，再利用OpenPose以及本研究開發出的鼻心物件演算法，針對五官遮蓋的進行細部判斷，進行誤判修正，最後證實出本算法能篩出72%被神經網路YOLO誤判為有戴好口罩的資料，可顯著提升口罩配戴辨識精準度，減少形成防疫破口的機會。

強化學習為當前AI領域的熱門話題，其特點是在環境的獎勵與懲罰下，進行學習。強化學習雖然較為困難，但其成功的項目都非常有名，其中最著名的例子有: AlphoGo、AlphaZero等等。 深度強化學習(DRL)是深度學習與強化學習的結合體，本專題透過DRL實現近端優化策略演算法，來使BipedalWalker環境中的二足模型學會行走，並調適超參數與神經網路來讓模型訓練擁有更好的結果。 經過實驗後發現，適當的降低獎勵折扣衰減率能有效的提升學習速度以及學習上限，同時可以避免分數落差過大導致的Dead relu問題。最終的結果能讓平均分數達到301分，成功達成了BipedalWalker環境要求(平均分數>=300分)。 為了使智能體擁有更好的探索能力，本專題加入了ICM(Intrinsic Curiosity Module)，成功提升了最終的平均分數至315分，將不摔倒的機率提升至99%，最高分數則到了320分，使得雙足模型能以更快的速度向前移動並保持穩定。

本研究利用NASA的SRTM 1 Arc-Second資料集來收集全球各地的地形高度圖(heightmap)，也利用MapTiler網站收集相對應的衛星空照圖。利用這些收集的圖像，訓練我們建構的VAE-pix2pix模型。VAE-pix2pix為Variational Autoencoder (VAE)及pix2pix (為一個Conditional Generative Adversarial Network)結合的模型，能將人工繪製的高度圖加上真實山脈應有的細節(包含尖銳的山脊、山壁上的紋路、連續的河流網路等……)，以生成出相對應的擬真衛星空照圖。相較於原pix2pix模型，VAE-pix2pix所生成的高度圖及衛星空照圖會更接近於真實世界的地形高度圖及衛星空照圖，同時VAE-pix2pix模型也可以透過改變latent code的數值來生成出不同風格的高度圖及空照圖，如地貌的顏色或雪線的高度等，這些都增加模型生成圖像的多樣性。為了使我們建構的模型能廣泛的被應用，我們在Unity上開發了Unity客戶端，其生成的mesh可以讓使用者直接應用於遊戲的場景，簡化了生成擬真山脈模型的任務。

In 2016, the population of people over the age of 65 in Macau was 11.2%. This means that Macau has already become an aging society. As such, more younger generations are needed to look after the elderly. According to statistics, falls are the primary cause of injury or death for the elderly over 65 years old. About 30% of the elderly over 65 years old fall every year. Along with the increase in the elderly population, it is urgent to find a fast and effective way to ensure the safety of the elderly. As there is a lot more danger besides falling in an elderly life, we aim to build a robot collocated with its danger detection system to ensure the safety of the elderly at home. The reason we decided to use robots is that we want to have larger flexibility and mobility, for example, we can give elderly rescue materials when they need help. Moreover, more home robots will be used in the future, they can just apply our system to theirs and ensure the safety of elderlies. In this research, we mainly used cameras with the Openpose model to detect dangers such as falling, potential human action danger, and environmental danger. Innovative ways are used to detect fall action, collocated with our home robot, it is a foreseeing project that could ensure the safety of the elderly in a home environment.

To this day, honeypots have proven effective in trapping attackers, misdirecting their malicious intents, and occasionally studying their tactics and tools without compromising any important production systems in the process. However, the lack of human network activity produced on honeypots have become potential giveaways of its true identity. Upon discovery, a honeypot is rendered much less useful. Apart from its steathiness, the enticing quality of a honeypot is also paramount in luring in attackers. Therefore, it is not surprising that there has always been ongoing work in these two aspects of honeypot development. While there are a great variety of network activities that potentially make a honeypot look realistic, web-surfing activities form a huge portion because it is a social norm and enables the honeypot to be detected by the attacker on the network as an active system. Hence, this paper describes our work in creating an intelligent honeypot which has learnt from human surfing behaviours to simulate similar activity, using a novel technique of “task-based learning”. A survey was conducted to evaluate how realistic the logs from our solution was relative to that of random browsing as well as real humans and the results show that the intelligent web-surfer managed to perform relatively well, achieving results close to that of real humans. In closing, an implementation of the proposed system would lead more attackers to believe that they have stumbled upon real production systems, helping threat hunters be on step ahead of attackers.

Abstract Computer vision is the science that aims to visually sense the world around it. It is concerned with the automatic extraction, analysis, and understanding of useful information from a single image or a sequence of images. Also, involved in the development of a theoretical and algorithmic basis to achieve automatic visual understanding. According to WHO, there are about 15% of the world population are suffering from physical disability, who are challenging to unemployment, difficulties of interacting with others and education. In this study, the researcher aims to help physically disabled people to use a computer. To achieve the purpose of the study the researcher analyzed the movement of the face and the tongue. This is done by using Python language to develop this program. The program analyses the range of colors, each color has a specific area on a mask the user wears, through an attached camera to the computer, the program detects changes of colors happens when the head or the tongue of the user moves. Once a change has been detected, the program calculates the movement occurred through the difference between color ranges to set the new location of the cursor on the screen, detect and set actions based on the user’s choice for right or left click or if the user has chosen to exit the mask mode and shut down the computer. From analyzing the results, the researcher concluded that by using computer vision and python libraries, it is possible to help the physically disabled in using computers. Keywords: Computer Vision, Python

Matheus Henrique Schmokel
matheus.schmokel@gmail.com

The Information Technology (IT) area has shown great growth in recent years, even with the economic recession that Brazil has been through and the impact of the coronavirus pandemic. It is estimated that by 2024 the area will have a deficit of more than 290 thousand professionals. However, companies still face other difficulties in hiring, especially people who are looking for their first job in the Information Technology area. Most part of these difficulties are: lack of qualified manpower and high prerequisites to fill internship or junior positions. As a result, the objective of this project is: to develop a platform that connects people who seek guidance, improvement or professional relocation in the Information Technology area with professionals that already have the experience they are seeking. The first step was a research and analysis of similar platforms in the market, whose proposal involves mentoring or professional connections, and it concluded that there are no services that fully meet the project’s proposal. In the second step, a research was done about mobile development, highlighting Flutter and Firebase platform. The third step defined the application’s features, such as suggestion of users and mentors, search for users, become a mentor, private chat, video calls, Portuguese and English languages, light and dark themes and profile customization. The suggestion of users and mentors is done by a match with the registered users, relating their areas of work (where the user has experience) and the areas of interest of each one. For the coding of the project, Flutter and Firebase technologies were used. To design the app, it followed Material Design specifications. For testing and distribution, the app was published on Play Store, Google’s Android application platform. The tests were performed by both the researcher and a selected group of users to verify if the functionalities were in accordance to what was defined in the beginning of the project. Perceiving the correct functioning of the application, the project achieved the proposed objective. In addition, it expanded its reach area, because it is possible to find users and mentors from any other area of the market.

Nowadays, all blind people are able to read any type of text thanks to The Braille system, which is a system of reading by touch. The Braille system provides a good and important help to blind people but it’s not enough. They try to integrate into society, try to read and write the same as those blessed with sight. However, only 18% of people can actually write in Braille and doing so takes a lot of effort. While Braille techniques have existed for many years now, there isn’t one that takes advantage of the comfort and easiness of modern technology. To be fair and give more chances to the blind people, and with all the available technologies nowadays, we must find solutions and innovative ideas to fulfill this objective. And this is how our project was born. To be honest, it’s not just a project, for us it’s a dream. Our focus in this project is creating an automatic machine based on transforming our language to their language and transforming this new invention to assist the blind in writing whole texts through speaking. It will be easier than the traditional way. This part of the project will help us print Braille on paper automatically in the cheapest, fastest and easiest way, unlike the traditional method that not only takes a lot of time but also needs a professional. Our printer will only cost approximately 250$ compared to regular printers that cost 6500$! the principles of the code of the speech recognition are voice pickup in all languages as first step, second speech recognition, then speech translation, after that speech analysis and finally translation into Braille and printing it. As for the future, we will work on face, object and image recognition as a scanner to make the visually impaired people read the name of the object and recognize people’s faces like anyone else.