* Accountable for performance analysis and improvement of algorithms in many projects like computer vision, OCR, object detection/segmentation, face recognition, and sign language recognition. * Research and implement appropriate deep learning algorithms and tools and develop applications according to requirements. - Successfully reduced errors in the defect detection dataset resulting from inaccurate annotations by 75%+ using a corrective learning process that repeatedly checks prediction results and corrects inaccurate annotations. - Reduced about 50% of recognition error compared to the SOTA models for sign language recognition by devising an effective model design using 1D convolution. Model training was done in minutes without a GPU.

更多

* Designed a way to prevent forks and dramatically speed up the transaction process by defining nodes that monitor failures and hacks using AI technology in a blockchain system. - Functionality design that detects potential forks and system failures using machine learning techniques. - Functionality design that recovers blockchains selectively and automatically.

更多

* 기계학습, 컴퓨터비젼, 추천 알고리즘 연구 * 소프트워어학과 학과장

* Participated in research activities associated with computer vision including face recognition, and object detection. - Successfully reduced 70% of face recognition errors using clustering algorithms to prevent an unregistered face from being classified as the closest registered face. - Used a face verification model to distinguish similar faces well, and the embedding vector obtained after the convolutional layer was used for face registration.

更多

* 기계학습 기반 IDS 연구