SOSCON 2019

I would like to introduce you to the people of Keras Korea. Let’s talk about your interests, your jobs, and your goals for the future.

PresenterTaeyoung Kim

To help people learn about deep learning by using Keras, a simple and intuitive deep learning library, I have written a book titled Python Deep Learning Keras with Blocks. I also run 'Kim Taeyoung’s Blog about Keras' and regularly post in communities such as 'Keras Korea' and 'Kaggle Korea'. With members of the 'Reinforcement Learning Korea’s Alpha-Omok Project,’ I have established services that allow the public to familiarize themselves with the AlphaGo model. Currently, I am the general director of technology at Interspace Co., Ltd., actively taking part in an R&D project to apply AI to various fields based on the company’s motto, “Deep learning from the Sun to the cells” and “Reinforcement learning from games to the universe”.

Nowadays, it is of great value to share technology with others. My presentation is about community activities through which people can grow together instead of growing alone. We will look into stories shared in the community and activities developed both online and offline by different people who are fascinated by the same field. I will also give you tips on how to communicate with others for mutual growth.

PresenterMijeong Jeon

I am an iOS developer who is open to anything. I like to learn new fields quietly by myself and share that information with others. I am not a computer major, so I went into programming somewhat belatedly, but I am enjoying every bit of being part of this world. It is always fun to talk about technology and communicate with different people within a community. Recently, I became fascinated by AI and machine learning whilst working at Kera Korea and as a Microsoft AI MVP.

Introducing the changes in TensorFlow 2.0 and the role of Keras in TensorFlow 2.0. In addition to using basic keras, we will also look at how to implement a custom model using Keras in TensorFlow.

PresenterHaesun Park

ML GDE (Machine Learning Google Developer Expert); I majored in mechanical engineering, but after graduation, read and wrote code. I'm exploring the boundaries between software and science interestingly, contributing to the translation of TensorFlow documents. I wrote a book titled "Do it! Deep Learning", and translated several books including “Python Machine Learning”, “Deep Learning with Python”, “Introdcution to Machine Learning with Python”, “Hands-On Machine Learning with Scikit-Learn and TensorFlow”.

Blog: https://tensorflow.blog
Github: https://github.com/rickiepark
E-mail: haesunrpark@gmail.com

As the Chatbot system has gained more popularity lately, many companies have attempted to establish their own Chatbot systems. However, the abstract concept of artificial intelligence has caused several misunderstandings. I would like to compare the existing text-based technology and the current AI (Deep Learning)-based technology and share the flow of natural language processing, which forms the core and acts as an example of the Chatbot system used for business. I would also like to introduce Keracon that can currently be found in the Open Source Contributhon.

PresenterSeulki Kim

Having majored in Electronics and Computer Engineering at Hanyang University, I began my career as an S/W engineer, which led to me becoming an M/L engineer today. I took part in a project establishing the Chatbot system for natural language processing and developing an AI cover letter evaluation solution that uses text-based technology. Currently, I am a mentor for the Keracon Contributhon.

The use of deep neural networks has expanded from Cloud or previous high-performance devices to the On-Device AI of mobile/house appliances. On-Device AI is greatly efficient especially if a massive Online Data Stream generated directly from the Device requires low latency processing, if the expense of the Cloud becomes a problem, or if you are sensitive to personal data protection and security issues. Lately, there have been ideas about establishing On-Device AI in a Data Stream Pipeline form to set up an AI Application that satisfies all of the above requirements. My presentation will be about NNStreamer, which is a Pipeline Framework for the neural network and can be commonly used in Tizen, Android, Ubuntu, Yocto, and the macOS environment. Then I will give an introduction to how to use NNStreamer to easily and effectively develop an application.

PresenterMyeongju Ham

I majored in computer science at the University of Illinois at Urbana-Champaign. I have been working at Samsung Electronics since 2009 and have developed system S/W with Kernel from Tizen. Afterwards, my team and I moved on to improve the Tizen app and the platform development environment. Recently, I have become a member of an automatic driving project and am currently developing the NPU System SW and NNStreamer (neural network pipeline) for On-Device AI. For my Open Source activities I proposed the devfreq and extcon frameworks for the Linux Kernel as a maintainer as well as a founder of NNStreamer. I was also a committer for several projects, including NET Runtime and Tensorflow.

It is okay to feel helpless at first no matter how much you want to contribute to Open Source. Keras Korea has prepared various projects such as Contributhon and the Korean translations of official Keras documents to help beginners take their first steps as Open Source developers. While striving to spread Open Source culture and contributing to Open Source projects so far from Keras Korea, I have collected ideas and opinions for this presentation to talk about how to approach the system and what information needs to be obtained in advance.

PresenterYounjoon Chung

After majoring in computer engineering at Sogang University, I became a member of the Microsoft Student Partners and held many deep learning workshops and seminars. Currently, I am working at an IT firm in the R&D department for computer vision. I have also engaged as a contributor to various Open Source libraries such as Keras and matplotlib, and I am in charge of the Korean translation project of the official Keras document from Keras Korea. I am also interested in expanding Open Source culture.

Keras is a deep learning library that is welcomed by many people because even beginners can use it with ease. However, a lot of them have trouble adding other functions or editing the layer of neural networks to their taste. We receive similar questions in open chat rooms. We plan to give an introduction to various functions that are useful but are still relatively unknown or which users have asked about so that users can feel free to use Keras and enjoy it.

PresenterJunwon Hwang

I am currently majoring in software at Sungkyunkwan University and serve as part of a Skilled Industrial Personnel team. I also take care of the open chat rooms as a Keras Korea staff member, and I give presentations at seminars or share opinions in various deep learning communities. I am interested in deep learning technology research and the development of applications overall, and I wish to provide useful tools and services to people.

How should you begin studying about data? You found a platform called Kaggle, but you are new to everything about it. You have heard of Keras, PyTorch, and TensorFlow, but you do not know where to begin. People say Keras is a piece of cake, but not to you. Is there a way to be good at Kaggle? Let me show you how the beginners can familiarize themselves with Kaggle through Keras. There are useful tips and information needed for Kaggle as well!

PresenterTaejin Kim

After graduating from university, I became so addicted to Kaggle that I refused to get a job and instead joined the Kaggle community. I believed that all of the studies I had done in school would only be completed outside of campus, which led me to participate in several school contests and club activities for business start-ups. I am indeed a bit too much into anything other than studying, but I do think that all of that experiences helped me grow! Currently, I am a staff member at Kaggle Korea Group.

Object Detection, a system that analyzes objects included those in video and image contents to find useful information or that provides a new user experience, is one of the most popular fields of computer vision in the present day. It boasts of high utility and popularity, so various APIs are being created to facilitate the use of Object Detection systems. However, there are difficulties in tuning the opened API to the right purposes or applying the actual data. All this led to the 2019 Open SW Contributhon Keras Community Project! Here is the Objective Detection Platform YOLK based on Keras.

PresenterJunghyun Park

My encounter with AI began with the Drone Detection system project based on Machine Learning at Purdue University, with which I had so much fun. I am currently working as a Machine Learning Engineer at kakaoVX, specializing in computer visions. Now I am taking part in an R&D project in the Human Activity Recognition field. My goal is to provide services targeting many people, and anything exciting will catch my attention! I am also working with great people as a Keras Korea staff member.

What would it be like to be able to identify oneself without typing in IDs or passwords? Let’s find out how AI can be used to create a digital identity and let's come up with an example of mobile signature recognition with Keras.

PresenterHyungsuk Kang

My fascination towards machine learning and AI has led me to major in Udacity machine learning nano degree and to create a document reader for the visually-impaired as a Google hack fair finalist. After learning the importance of collecting and processing data, I invented Naturecoin, which is a system that rewards users in return for the image data and consumer data about waste when throwing away recyclables in order to resolve environmental issues. This invention saw me win the GIX Innovation Challenge and become a Citypreneur finalist. I began to take interest in the decentralized identity (DID) and data privacy, and I am currently conducting an R&D project on bridge and identity verification from SpeckleOS to protect data portability and privacy between blockchains within the Polkadot Network (an interchain project).

Little Big Data is a community that irregularly holds meetups with the slogan, "Big Data is not only for Big Companies." We share stories about data with people from various backgrounds.
https://www.facebook.com/groups/littlebigdata/

Play with Data is a festival for everyone who deals with data. While interest in data is increasing day by day, it is hard to find people who can "play" with the data around us. Play with Data encourages us to think about how data users can "play" with data and hopes to share these idea with you.

PresenterYurim Jin

Leader of the Little Big Data Community

PresenterMingu Kang

Leader of the Play with Data Community

I heard that the Community Room of SOSCON was looking for people who would make a 25-minute presentation. But I struggled to find a proper topic for the theme titled, "How to create a side project with data." But I really wanted to make a presentation. So, I hurriedly started a side project. How much of a project called, "Automated Trading of Cryptocurrency Side Project," could be done by the presentation day if I started it from scratch? 1. When you see others who built a side project, you will say, "Wow," but when you do it yourself, it does not happen so easily. So how much time did they actually spend on the project? So here I will show you the reality of the process of building a project for the day of the presentation. 2. I will introduce the entire process from collecting and preprocessing data to automated trading using AI (plan).

PresenterChioh Song

Chio Song started cryptocurrency trading later than others while cryptocurrency was booming, but he earned 20 times his original investment. Alas in the end he spent everything he earned but is now happily working as a machine learning engineer in the healthcare industry.

People who are new to LOL or who are not familiar with it often buy items recommended by LOL or simply those items they want to buy. However, please check the items LOL recommends. LOL recommends specific items according to what type of character a user has, such as Tank and DPS. In particular, if you have poor knowledge of items on sale and agonize over which item to buy, you are joining the game late. Moreover, no one knows where a small snowball will bounce. Therefore, we try to recommend items using an algorithm that recommends items considering, for example, the champion of an opposing team, rune/masteries and combinations by time, such as the beginning of a game, the middle of the game and the latter part of it. I will fill 25 minutes in an upgraded way compared to the existing video that I’ve created.

Reference Video : https://www.youtube.com/watch?v=yUWnDQiybIA&t=3s

PresenterJaehwa Lee

I am AIRIM and I am a YouTuber on the subject of AI. We are living in an era where anyone can deal with data even though they are not a data analyst. However, entry barriers are so high. Therefore, I explain how to deal with data in a very easy and fun way to people and tell them it is not difficult. I make YouTube videos for those who have no idea about data analysis and for those who want to analyze data.

I will share the processes of and review a project titled, "Data Analysis of My Brunch Subscribers," that I started after learning Python this year. For the project, I carried out web crawling and data visualization, and as a result, I learned a simple data insight. In this presentation, I will tell those of you who are interested in data utilization and analysis that anyone can find an answer to their interest and curiosity by using data and, furthermore, utilize data and apply it to work.

PresenterWonyang Kang

Working in the field of data visualization, Wonyang Kang tells us about various types of data visualization, such as planning, content production and education of data visualization.

I will tell you how I started studying data science as a high school student and explain how to easily play around with data based on my own experience. After that, we will have time to be familiar with data through data visualization and then predict the stock price of Samsung Electronics based on its past stock price data, and after that, we will discuss whether the result of this activity is relevant. Lastly, I will introduce Kaggle through which anyone can easily access data science.

PresenterHyunwoo Kim

Hyunwoo Kim is currently teaching juniors at an AI study group called, "Emotion," and studying at an intensive AI study group called, "Evolution," at Sunrin Internet High School. Recently, he made a presentation titled, "AI: How do you learn it?" at the Data Science Youth Conference.

It started with the question, "Our soccer team members are still of a young age (in their 20s to 30s) and they are quite good, but why do we always lose?" I am not a very competitive person, but I really wanted to win a soccer game. I collected data, including unexpected elements, such as whether or not the team's ace player drank alcohol the night before a match. It is a very simple model, but thanks to this, it feels nice because we can choose a team that we can defeat and play a game against them.

PresenterSanghoon Shim

A Data Anaylst who Loves Soccer

Introducing Side Project and Sharing Entire Flow and Results (YouTube Public API + Python + NodeXL)

PresenterSungsan Ro

Seongsan Noh is in charge of DataLab at Sandbox. He is a very curious person and enjoys trying new things.

It will be a fun networking experiences for presenters to analyize data.

With the advent of the Fourth Industrial Revolution, technological advancement and the rapid modularity of robot hardware, we can easily imagine that robots will be used in our daily lives in the near future. Lots of engineers around the world are participating in the advancement of robot technologies equipped with autonomous driving technologies and AI, and in South Korea, many related online communities are being created and actively operated. I will introduce some of the online communities related to ROS in South Korea, focusing on "OROCA," an online community created in 2012.

PresenterKyunam Choi

As a principal research engineer at the healthcare research center at Gachon University, Gyunam Choi is currently researching and developing healthcare robots and is active in a Naver online community about robots called, "OROCA." Also, he is focusing on popularizing ROS and activating Meetup.

I will introduce how ordinary people who are interested in robots managed to learn ROS and develop a mobile robot and participate in the Korea Intelligent Robot Contest. Also, I will introduce how these ordinary people used ROS and open source in terms of robot development, what difficulties they faced and how they overcame these challenges.

PresenterJihyeon Ha

Jihyeon Ha majored in mechanical engineering and electrical and electronic engineering at Sungkyunkwan University and is now in charge of the design of electric apparatus/SW. Jihyeon Ha is interested in robot software and applications.

I will introduce the development process of the KT hotel robot, which is expected to be commercialized in the Novotel Ambassador Seoul Dongdaemun Hotel. I will share how I built a service robot system for commercialization and what difficulties I faced during the process.

PresenterTaeho Kang

Taeho Gang majored in electronic engineering at Hanyang University and completed his master's degree at a robot control lab. He is now researching and developing indoor autonomous driving technologies for mobile robots based on ROS at the KT Institute of Convergence Technology.

It is a program where participants suggest ideas for creative robots that help human beings, using open source. The finalists will give a group presentation and be judged, and the suggested robots will be exhibited.