일시 및 장소: 2019년 9월20일 10:30, RA204 강사: 임성훈 교수, DGIST 정보통신융합전공제목: Incorporating Geometric Knowledge into a Deep Neural Network 부제: - to Explore Synergies in 3D Computer Vision -초록: Recently, many smartphone manufacturers are actively studying 3D reconstruction to enable various AR / VR and photography applications. To support the features, they have developed a new multi-camera system mounted on mobile phones and it has become a new trend in the mobile industry. However, the performance of depth estimation heavily depends on the scene environment, such as texture-less region and illumination changes. To address these photometric issues, I proposed a robust 3D reconstruction method for such challenging conditions. I achieved this by investigating the advantages of two different research field, traditional 3D geometry and recent breakthrough in deep learning. I mainly studied what knowledge of 3D geometry to incorporate into a deep neural network, and how to integrate them to explore synergies in 3D computer vision. In this talk, I will share my two different attempts and results of the incorporation. Part of the content of this talk was published in CVPR2018 and ICLR2019.
1. 주제: Logic Compatible Flash Memory Based Non-Volatile Neural Network Accelerator2. 연사: 송승환 박사 (ANAFLASH, USA)3. 일시: 2019년 7월 16일 10:30 ~ 12:304. 장소: 서강대학교 R관 715호초록Power efficient edge computing with some degree of Artificial Intelligence (AI) is a massive trend. The society is moving to include AI feature virtually in every edge device such as intelligent home product, smart mobile, security monitor, wireless sensor, self-driving car to name a few. A cost-effective deep neural network (DNN)accelerator is a key building block for such edge devices. Existing solution for the DNN HW typically requires off-chip access to retrieve neural network parameters from external memories, incurring huge communication latency and power consumption. Alternative approach integrating the DNN engine in a special Non-Volatile Memory (NVM) process requires additional steps beyond the conventional logic CMOS process which is not cost-effective for medium density DNN engine in cost-sensitive edge devices. In this presentation, single-poly embedded flash memory implemented in a generic logic process is revisited for moderate density Non-Volatile DNNapplications. Then, the latest test chip result of the logic compatible flash based neural network accelerator featuring high energy efficiency and fine-grained analog programming scheme will be discussed.송승환 박사 약력Dr. Seung-hwan Song is a CTO and Co-Founder of ANAFLASH (www.anaflash.com), Inc. that develops logic compatible embedded flash memory based solution since 2017. Previously, he has held variousresearch and development positions at SK-HMS America, HGST (acquired by WD), Seagate, Qualcomm, Broadcom, and Samsung. He received the B.S. and M.S. degrees in electrical engineering from Seoul National University,in 2004 and 2006, and the Ph. D. degree in electrical engineering (with a minor in biomedical engineering) from University of Minnesota, in 2013. Dr Song is a principal investigator of the company’s SBIR awardsreceived from NSF and DoD Air Force, and has 45+ US patents granted around SSD, NVM, and Flash memory. He is a recipient of a best paper award at IEEE ICC 2016 and a low power design contest award at IEEE ISLPED 2012.세미나 문의: 전자공학과 최우영 교수 (email@example.com)
Sogang Included in Preliminary Selectionfor Phase Four of Brain Korea 21 Project (BK21) The Ministry of Education and National Research Foundation of Korea (NRF) have announced the preliminary selection results for Phase Four of the Brain Korea 21 project (BK 21), which supports universities with research funds and other aid for stabilized academic research and technological development in new industries. Sogang University has been honored with the selection of three Educational Research Centers and five Educational Research Teams, as follows: Educational Research Centers:△ Fostering Talents in Future Brain-copied Intelligent Semiconductor Innovation (7 professors from Department of Electronic Engineering)△ Mobile Blockchain-based Intelligent Healthcare Solution for Areas of Disasters/Emergencies (7 professors from Department of Electronic Engineering)△ Fostering World Highest-standard Practical Big Data Convergence (12 professors from Department of Computer Science and Engineering)Educational Research Teams:△ Roots-based Politics (6 professors from Department of Political Science)△ Anti-stress Bio-molecular Function (7 professors from Department of Life Sciences)△ Molecular Control-based Chemical and Biochemical Engineering Process (7 professors from Department of Chemical and Biomolecular Engineering)△ Fostering Professionals in K-MENA (4 professors from Department of History)△ Smart Mechanical Technology for Social Value Creation (7 professors from Department of Mechanical Engineering) Accordingly, Sogang will be provided research funds for seven years (beginning in September 2020) in three stages - totaling approximately 80% more than the funds received for Phase Three of BK21 (2013 to 2020) - indicating a more stabilized fostering of graduate and doctoral researchers to lead the Fourth Industrial Revolution. President Park Jong Gou noted, "This preliminary selection for Phase Four of BK21 is the result of Sogang members' efforts and cooperation," and expressed his hope that this will serve as momentum for more students to concentrate on research and grow into convergent human resources that will be much needed for the future society."
The joint research team led by Prof. Song Tai-Kyong and prof. Chang Jin-Ho from the Department of Electronic Engineering at Sogang University developed a world-first convergence imaging technology to innovatively improve the accuracy of sentinel lymph node biopsy guidance, which is used to detect breast carcinoma metastasis. Through preclinical testing, the effects and availability were technically proven for clinical usage. The outcome of the research was posted in the March 22nd issue of Scientific Reports, an affiliated journal of Nature (https://www.nature.com/articles/srep45008).If this equipment for convergence imaging diagnosis, which is being developed at this time, is commercialized, the limitations will be overcome for treatment and diagnosis using a single imaging device. For instance, preclinical testing has demonstrated that the precise identification of the sentinel lymph node (the primary lymph node that a cancer cell passes through in the case of metastasis) and the accuracy of biopsy guidance can be enhanced.Prof. Song said, “Thanks to this technology, ultrasonic and photo-acoustic images are now available ahead of biopsy guidance, and doctors can reconfirm during the biopsy procedure whether the lymph node to be removed is the signal lymphatic gland or not, using the fluorescent images.” He added, “This enables fast and accurate biopsy procedures and a minimum incision on the part to be used for the biopsy. It is expected that more accurate decisions on carcinoma metastasis and biopsy procedures will minimize the side effects and complications that a patient must endure.”It is also projected that the technology will be utilized for clinical applications in a variety of ways, such as through convergence image guidance surgeries aimed at enhancing the accuracy of cancer tumor removal.
Sogang University Signs MOU with LG Electronics for Nurturing Smart Convergence Majors Sogang University and LG Electronics signed an MOU for opening and operating a track for nurturing a smart convergence-specialized department. Based on the MOU, the Engineering College at Sogang will open a track for nurturing R&D human resources for LG Electronics (LG Electronics Track) to foster specialized human resources in the fields of IoT, robotics, AI, and autonomous navigation vehicles at Sogang University. From June 2017 to May 2020, five to ten undergraduates and graduate school students will be selected annually for the LG Electronics Track. These students will be supported in their two-year graduate school courses through the industry-academy scholarship program, and junior-year students will be given an additional scholarship for their senior year. Furthermore, all selected students will have the option of working at LG Electronics after their graduation. Student selection for 2017 is slated to run from August to September.
Prof. Jang Jin-ho’s Research Team at Sogang Institute Advanced TechnologySelected for ICT Creative Task at Samsung Science & Technology Projectto Receive Total of KRW 1.4 Billion for 4 Years Prof. Jang Jin-ho’s research team at Sogang Institute Advanced Technology (SIAT) at Sogang University has finally been selected for an ICT creative task at Samsung Science & Technology Project that provides support in the latter half of 2017. The Samsung Science & Technology Project which started in 2013 is a program to select excellent tasks three times a year in three areas of research of fundamental science, materials technology, and ICT, thereby supporting and nurturing future technologies in the Republic of Korea by providing a total fund of KRW 1 trillion 500 billion for 10 years. For the latter half of 2017, 14 tasks for the fundamental science and 23 tasks for the materials technology and ICT have been selected including the task of “development of ultrasound-assisted confocal microscopy for deep tissue imaging” by Prof. Jang Jin-ho, and a total of KRW 53.1 billion will be provided. The selected task of Prof. Jang is to develop an original technology to settle the issue of a low image depth which is an inborn limitation of the conventional confocal microscope, aiming to develop a technology to increase an image depth of a confocal microscope without a lowered image resolution by using ultrasound wave. A confocal microscope is a device that focuses and irradiates light to a living tissue and that uses a reflected light signal, thereby providing structure information of a microtissue such as a cell with a high-resolution, high-contrast image. Currently, the confocal microscope is the most broadly used as a three-dimensional high-resolution imaging technique for life science researches and disease diagnosis including images of cancer, brain tissues, etc. However, though a high-resolution, high-contrast image provided by the confocal microscope is available only at a depth in which light can be focused, the depth for focusing light is only hundreds of micrometers due to optical scattering in a living tissue. Thus, it is difficult to image a tissue in a deeper part. To resolve the issue, Prof. Jang Jin-ho’s research team will develop an original technology for a confocal microscope, a technique that is expected to increase a depth of light penetration by temporarily changing characteristics of a medium to minimize optical scattering in the medium in which the light proceeds, using an ultrasound energy recently developed by the team. The proposed technology is an original technology with a new concept, which can not only remarkably broaden an application area of a confocal microscope that has limitations triggered by a low image depth but also expand the application area of the conventional optical image and treatment technologies that have limitations in use because of a low depth of light focusing. Thus, the technology has a very great impact to satisfy unmet clinical needs. If the technology is commercialized, one can easily enter into and expand a share in the existing optical image market, an area that is mostly occupied by overseas technological leading companies, thereby creating a new market and leading the relevant industries in the whole world.
Dept. of Electronic Engineering, Won the Bid for ‘2018 University ICT Research Center Project of the MSIT- Incubating Human Resources for Developing Intellectual Semiconductor for the 4th Industrial Revolution Era -- Aims to Foster 111 Masters and Doctors for the Next 4 Years - Sogang University was selected as the University ICT Research Center for incubating talents in the ICT sector by the Ministry of Science and ICT on June 10. The University ICT research center project is a representative project in the ICT sector that has supported the university-industry cooperation and the incubation of high quality human resources through technology development at universities from 2000. This year, eleven universities including Sogang won the bid in such strategic industries for the 4th industrial revolution as the intellectual semiconductor in which the industry empowers the core technology for the 4th industrial revolution, data, network, artificial intelligence, and software. The selected institutes will be given the government support of approximately 3 billion Korean won over the next four years and additionally, they will be supported by fund to respond to the private companies. Prof. Lee Seung-hoon of the Department of Electronic Engineering is the chief of the center and 17 professors, including prof. Kang Seokju, Kim Kwangsu, Beom Jinwook, Ahn Gilcho, Yoon Kwangsuk, and Choi Wooyoung from seven universities and 13 enterprises, including Samsung Electronics, SK Hynix, and Silicon Works, participate in the Intelligent Semiconductor Design Research Center (ISDRC). The center aims to foster high quality human resources of hardware-software convergence who will take the initiative in this era of the 4th industrial revolution based on developing designing technology for super intellectual, super energy saving, and highly credible intellectual semiconductor platform and device-circuit. The ISDRC plans both to foster researchers, 98 master’s degree holders and 13 doctoral degree holders, over the next 4 years, and to conduct various cooperation programs of re-training of engineers working for the participating enterprises, when it is deemed required. Throughout the activities, the ISDRC will play the critical role of the pivot for academy-industry cooperation and the incubator for talents in order to promote a longer-term growth of the domestic intellectual semiconductor sector.