Adversarial Strategies for Anomaly Detection and Image Restoration - SIAM-IS Virtual Seminar

Explore adversarial strategies for out-of-distribution detection and inverse problems in imaging during this virtual seminar talk from the Tenth Imaging & Inverse Problems (IMAGINE) OneWorld SIAM-IS series. Join speaker Coloma Ballester from Universitat Pompeu Fabra as she delves into innovative approaches in anomaly detection and image restoration. Discover a novel method for unsupervised learning of normal data distributions using GAN strategies, and learn about a new anomaly detector that leverages historical data from normal generators. Gain insights into efficient and versatile anomaly detection techniques applicable across various data modalities. Additionally, examine adversarial approaches for tackling image restoration challenges, including inpainting and colorization. This hour-long presentation offers valuable knowledge for researchers and practitioners in the fields of imaging, inverse problems, and machine learning.