Awards & Press

KAIST EE Presented Cutting-edge AI Research Results At CVPR 2019
CVPR is the premier computer vision conference in the world. The quality and impact of instituion's research activity in computer vision and artificial intelligence are often measured by the number of papers accepted to this conference. KAIST EE has been very prolific in this sense. At 2019 CVPR alone, KAIST EE researchers have published 12 papers, becoming one of the most productive institutions of the world in computer vision and artificial intelligence research. These papers can be found below:   Deep Blind Video Decaptioning by Temporal Aggregation and Recurrence Dahun Kim, Sanghyun Woo, Joon-Young Lee, In So Kweon Deep Video Inpainting Dahun Kim, Sanghyun Woo, Joon-Young Lee, In So Kweon Dense Relational Captioning: Triple-Stream Networks for Relationship-Based Captioning Dong-Jin Kim, Jinsoo Choi, Tae-Hyun Oh, In So Kweon Learning Loss for Active Learning Donggeun Yoo, In So Kweon Variational Prototyping-Encoder: One-Shot Learning with Prototypical Images Junsik Kim, Tae-Hyun Oh, Seokju Lee, Fei Pan, In So Kweon dge-Labeling Graph Neural Network for Few-shot Learning  Jongmin Kim, Taesup Kim, Sungwoong Kim, Chang D. Yoo Progressive Attention Memory Network for Movie Story Question Answering Junyeong Kim, Minuk Ma, Kyungsu Kim, Sungjin Kim, Chang D. Yoo Diversify and Match: A Domain Adaptive Representation Learning Paradigm for Object Detection  Taekyung Kim, Minki Jeong, Seunghyeon Kim, Seokeon Choi, Changick Kim Learning Not to Learn: Training Deep Neural Networks with Biased Data Byungju Kim, Hyunwoo Kim, Kyungsu Kim, Sungjin Kim, Junmo Kim RL-GAN-Net: A Reinforcement Learning Agent Controlled GAN Network for Real-Time Point Cloud Shape Completion  Muhammad Sarmad, Hyunjoo Jenny Lee, Young Min Kim Efficient Neural Network Compression  Hyeji Kim, Muhammad Umar Karim Khan, Chong-Min Kyung Variational Information Distillation for Knowledge Transfer  Sungsoo Ahn, Shell Xu Hu, Andreas Damianou, Neil D. Lawrence, Zhenwen Dai
KAIST EE Presented Cutting-edge AI Research Results at ICML 2019
ICML and NeurIPS are the world's most prestigious machine learning conferences. The quality and impact of institution's research in machine learning are often measured by the number of papers accepted at these conferences. KAIST EE has been very prolific in this sense. At 2019 ICML alone, KAIST EE researchers have published 9 papers, becoming one of the most productive institutions of the world in machine learning research. These papers can be found below:     TapNet: Neural Network Augmented with Task-Adaptive Projection for Few-Shot Learning Sung Whan Yoon, Jun Seo, and Jaekyun Moon  Dimension-Wise Importance Sampling Weight Clipping for Sample-Efficient Reinforcement Learning Seungyul Han and Youngchul Sung Weak Detection of Signal in the Spiked Wigner Model Hye Won Chung and Ji Oon Lee QTRAN: Learning to Factorize with Transformation for Cooperative Multi-Agent Reinforcement Learning Kyunghwan Son, Daewoo Kim, Wan Ju Kang, David Earl Hostallero and Yung Yi Learning What and Where to Transfer Yunhun Jang, Hankook Lee, Sung Ju Hwang, and Jinwoo Shin Training CNNs with Selective Allocation of Channels Jongheon Jeong and Jinwoo Shin Robust Inference via Generative Classifiers for Handling Noisy Labels Kimin Lee , Sukmin Yun, Kibok Lee, Honglak Lee, Bo Li, and Jinwoo Shin  Using Pre-Training Can Improve Model Robustness and Uncertainty Dan Hendrycks, Kimin Lee and Mantas Mazeika Spectral Approximate Inference Sejun Park, Eunho Yang, Se-Young Yun, and Jinwoo Shin
Professor Yongdae Kim, developed anti-drone technology for anti-terrorism
Our department’s laboratory (Prof. Yongdae Kim) has developed an anti-drone technology that can steal other drones by deceiving the location by using fake GPS signals. This technology can safely guide the drones to the desired location without any sudden change in direction at any emergency situation. Therefore, it can respond effectively to drones with the purpose of any terrorism. This technology was published on ‘ACM Transactions on Privacy and Security, TOPS’ on the 9th of April. As the industry of drone has developed, drones are utilized in various areas such as searching, rescuing, disaster prevention, response, delivery, reconnaissance. In contrast, concerns about private property, key facility intrusion, safety and security threats, and privacy invasion are growing. Therefore, the industry of detecting and preventing drones penetrating is rapidly growing. Currently, the anti-drone systems built in key facilities such as airports utilize electronic jamming signals, high-power lasers, or nets to neutralize drones. However, drones for terrors, which are armed with explosives or weapons, must be neutralized with securing a safe distance to minimize any damages. At the point where new anti-drone technology is needed, Professor Kim’s research team has developed new anti-drone technology that steals drones by tricking them with fake GPS signals. The research team analyzed the GPS safety mode of the drones made from major drone makers such as DJI and Parrot and made a classification system based on this. And they designed a drone abduction technique depending on the type of the drone. This classification system covers almost all the types of drone GPS safety modes and is universally applicable to any drones that use GPS regardless of model or manufacturer. The research team applied the new technology to four drones and have proven that the drones can be safely guided to the direction of intentional abduction within a small margin of error. Professor Kim said, “Conventional consumer drones seem to be safe from the fake signal GPS attacks due to equipped safety GPS mode, but most of them are able to detour because they detect GPS errors in a rudimentary manner. In particular, the new technology will be able to contribute to reducing the damage to the airline industry and the airport caused by illegal drone flight. The research team plans to commercialize the technology by applying to existing anti-drone technology by technology transfer.   <Anti-drone system developed by Professor Kim's research team>   <Link>  



No more events.