Hi! I'm Woojin Shin and I’m a master’s student in Artificial Intelligence at the University of Seoul.
Proposed BankTweak, a novel adversarial attack on multi-object tracking (MOT) that manipulates feature banks rather than object positions, overcoming the limitations of prior methods that are easily neutralized by Mahalanobis-threshold tuning and suffer from low stealth due to increased false positives and false negatives.
CF-DETR (Coarse-to-Fine DETR) adopts a two-stage detection strategy, first performing coarse localization and then refining key regions. By focusing computation where it matters most, it achieves higher efficiency while maintaining strong accuracy, making it suitable for real-time, resource-limited scenarios.
University of Seoul
Sep. 2024 ~ Feb. 2026
M.S. in Artificial Intelligence, On-Device Artificial Intelligence Lab
Supervisior: Prof. Hyeongboo Baek
Incheon National University
Mar. 2016 ~ Feb. 2023
B.A. in Economics with a Minor in Computer Engineering