Projects

R&D Projects

Ongoing

초실감 테라미디어를 위한 AV부호화 및 LF미디어 원천기술 개발

Start Date: 2021-01-01 End Date: 2028-12-31
Project Reference: 12149 Funding: Ministry of Science and ICT, Rep. of Korea (ETRI)
Duration: 12 months ~ Project cost: 52,531$ per year
Role: Principal Investigator
  The major purpose of this project is to develop the International Standardization of Ultra High-Definition Flat Video and Ultra-Capacity Cubic Video to Overcome the Compression Rate Limits of Existing Video Coding Technologies. Our team is investigating efficient deep learning algorithm to make more coding gain for inter mode prediction in the recent VVC standard ( https://jvet.hhi.fraunhofer.de/ ).

Development of profile measurement and nondestructive inspection system for cold forging product

Start Date: 2022-04-01 End Date: 2023-03-31
Project Reference: 12149 Funding: Ministry of SMEs and Startups, Rep. of Korea
Duration: 12 months ~ Project cost: 52,531$ per year
Role: Principal Investigator
  We are developing an automatic inspection system for cold forging structure shape measurement and non-destructive quality, and optimizing product shape measurement and non-conformity inspection algorithm for production aspect using machine vision and deep learning method.

Development of real-time intelligent object detection technology for drone captured image

Start Date: 2022-01-01 End Date: 2023-12-31
Project Reference: - Funding: Ministry of Science and ICT, Rep. of Korea (NRF)
Duration: 24 months Project cost: 63,130 $ per year
Role: Principal Investigator
  Deep learning networks for object detection have been actively studied so far, but one of the big problems that have not yet been solved is the detection and identification of small objects (about 32x32 pixels). In particular, in the case of drone captured images, there are many long-distance photographed objects, and it is very important that how the deep learning network extracts the small objects efficiently. This research aims to design and develop an effective deep learning network model for detecting missing people and small objects in high-resolution images taken by high performane drones.

Development of deep Learning based vision inspection system for bamboo toothbrush

Start Date: 2021-08-01 End Date: 2022-07-31
Project Reference: Funding: Ministry of SMEs and Startups, Rep. of Korea
Duration: 12 months Project cost: 52,531$
Role: General Research Director
  This project develops a technology that automatically detects four kinds of errors in the process of producing bamboo toothbrushes with artificial intelligence technology. Through this, it is intended to provide information on the type of defect and the location of the defect to the reprocessing process.

Evaluation of Speech Recognition on Foreign Language Education Services in Online Video Platform Environment

Start Date: 2021-07-01 End Date: 2022-04-30
Project Reference: Funding: Ministry of SMEs and Startups, Rep. of Korea
Duration: 8 months Project cost: 21,889$
Role: Principal Investigator
  This project will develop voice recognition pronunciation evaluation system technology optimized for online video platform environment in foreign language education services that help students learn English pronunciation on their own in an online video platform environment.
R&D Projects

Completed

AI-based Psychological Analysis Algorithm Using Image Recognition

Start Date: 2020-12-01 End Date: 2022-02-28
Project Reference: Seoul R&BD Program (CT200009) Funding: SBA, Seoul
Duration: 12 months Project cost: 87,556$
Role: General Research Director
  funded by the CT200009, Seoul Campus Town Technology R&D Project. In this project, we are devise an efficient scheme to analyze the hand drawings based on a hierarchical structure, which allows us to produce certain results. First, we extract some separated parts which is consisted of whole objects by object detection. With the extracted parts, we perform a classification task to define key attributes to analyze people's psychology. For the given image, the analyzed report is constructed based on the extracted key attributes.

Developments on Context Information Analysis Based Matching Technologies for Expanding the Market of Online Publications

Start Date: 2019-04-01 End Date: 2021-12-31
Project Reference: Funding: Ministry of Culture, Sports and Tourism (MCST) and from Korea Copyright Commission, Rep. of Korea
Duration: 33 months Project cost: 175,026$
Role: Principal Investigator
  This research project is being supported by Ministry of Culture, Sports and Tourism (MCST) and from Korea Copyright Commission in 20XX. This project aims at exracting text from electronic publications (PDF, EPUB, HTML, TXT, etc.), and applying artificial intelligence technology to the extracted text to analyze context information (person, relationship, emotion, time, space, event), and developing context information-based index and matching technologies to build a specialized creation (fiction, essay, script, poetry, etc.).

Development of SW for unidentification of self-driving image data

Start Date: 2020-09-25 End Date: 2021-02-28
Project Reference: Funding: Industrial Fundings by AI Networks Ltd.
Duration: 6 months Project cost: 19,687$
Role: General Research Director
  The purpose is to develop deep learning-based SW modules that detect various objects, especially objects related to personal information such as human faces and vehicle license plates, in driving situation.

Research on core technology for the next generation video coding based on visual perception

Start Date: 2017-11-01 End Date: 2020-06-31
Project Reference: Funding: Ministry of Education, Science and Technology, Rep. of Korea
Duration: 44 months Project cost: 192,610$
Role: General Research Director
  In this study, we study bit rate reduction algorithms that can efficiently encode with significantly increased bit information volumes in line with HEVC HDR/WCG. We also investigate a cognitive-based subjective quality improvement and bit rate reduction algorithms considering human cognitive mechanisms. Finally, we study high-efficiency coding techniques based on the Perceptual Video Coding framework that are suitable for next-generation ultra-quality content coding and fuse with human visual cognitive mechanisms.

Development of tire safety vision inspector using Deep Learning

Start Date: 2020-05-01 End Date: 2021-04-30
Project Reference: Funding: Ministry of SMEs and Startups, Rep. of Korea
Duration: 12 months Project cost: 78,759$
Role: Principal Investigator
  In this study, we developed automatic tire safety vision inspectors using deep learning techniques, to provide advanced customer service in retail shops.

Developments on extraction and classification of educational publication based on artificial intelligence technologies

Start Date: 2017-06-01 End Date: 2019-12-31
Project Reference: 2017-KCC-R001 Funding: Ministry of Culture, Sports and Tourism (MCST) and from Korea Copyright Commission, Rep. of Korea.
Duration: 31 months Project cost: 147,755$
Role: Principal Investigator
  The purpose of this project was a development of management technology to automate the copyright management procedure of educational contents using visual intelligence information technology and reduce unknown copyright of textbook works. In this project, our group has focused on efficient extraction of figures and their captions in textbooks.

Development of ICT-based smart machine tools and flexible automation systems

Start Date: 2015-12-01 End Date: 2019-07-31
Project Reference: 10060188 Funding: Ministry of Trade, Industry and Energy, Rep. of Korea
Duration: 45 months Project cost: 130,372$
Role: Principal Investigator
  The objective for this project was to build on Development of Machine PLC-based Smart Unmanned Operating System and Smart Machine Module for smart factory. Our team has developed data communication module and M2M communication based on the MTConnect standard for sensors in machines.

Development of a Smart Fitness System for Growing Kids by Using Multi-channel Wireless Communication

Start Date: 2016-03-01 End Date: 2017-12-31
Project Reference: KSIC Code 26329 Funding: Ministry of Trade, Industry and Energy, Rep. of Korea
Duration: 22 months Project cost: 260,602$
Role: Principal Investigator
  In this study, children's smart wearable devices and interlinked platforms have been designed and develpoed for up to 12 children to use at the same time, and additionally develop wearable device caps, stickers, and accessories using various characters to encourage children to actively perform fitness content. Also our group has suggested a new hand gesture recognition algorithm based on Gated Recurrent Unit (GRU) and bluetooth motion sensors. You can refer to deepGesture: Deep learning-based gesture recognition scheme using motion sensors .

A Study on Deep Learning-Based Inter-Frame Coding Performance and Complexity Improvement

Start Date: 2020-07-01 End Date: 2020-11-30
Project Reference: Funding: ETRI (IITP), Rep. of Korea
Duration: 6 months Project cost: 34,746$
Role: Principal Investigator
  In this study, we analyzed the prediction technology between frames using deep neural network structures among video coding technologies and proposed a deep learning-based video coding structure for image quality improvement based on this.