Center for Advanced Intelligence Project

Date and Time:
December 18, 2024: 15:00 – 16:00 (JST)
Venue: Online and Open Space at the RIKEN AIP Nihonbashi office*
*Open Space is available to AIP researchers only

Title: Reliable Model Predictions in the Open World Through the Lens of OOD Detection

Speaker: Yadan Luo (ARC DECRA Fellow and Senior Lecturer at the University of Queensland, Australia)

Abstract:
In the era of abundant and diverse data, the challenge in vision-based AI solutions transcends building deeper models. The critical focus now lies in ensuring robust and reliable model predictions when deployed in the unpredictable open world. Open-world scenarios introduce significant challenges, including distribution shifts in both data and labels, which can undermine the performance of classifiers and object detection systems. This talk explores our group’s recent works presented at ICLR ’24 and NeurIPS ’24, focusing on out-of-distribution (OOD) detection for classification and object detection tasks. It highlights novel approaches leveraging provable density modeling and unsupervised prompting to address these challenges, paving the way for more generalizable and trustworthy AI systems.

Bio:
Yadan Luo (https://luoyadan.github.io/) is an ARC DECRA Fellow and Senior Lecturer at the School of Electrical Engineering and Computer Science (EECS), University of Queensland (UQ), Australia. Her research focuses on enhancing the generalization capabilities of 2D and 3D vision models, with applications in autonomous driving and beyond. Yadan has served as the Publications Chair for ACM Multimedia, Short Paper Chair for The Web Conference (WWW), and Associate Editor for IEEE Transactions on Circuits and Systems for Video Technology (TCSVT). She has also contributed as an Area Chair for prestigious conferences such as CVPR, ICLR, and ACM Multimedia. Yadan’s contributions to the field have been recognized with numerous awards, including the Outstanding Service and Outstanding Area Chair Awards, Google PhD Fellowship, UQ Foundation Research Excellence Award, ICT Young Achiever Award from Women in Technology (WiT), and Best Student Paper and Best Demo Awards at ACM Multimedia 2023.