This CSAFE Center Wide Meeting Webinar was presented by Adam Kortylewski from the University of Basel in Switzerland on February 10, 2017.
The forensic analysis of impression evidence, such as fingerprints or shoeprints, plays a critical role in crime investigation. A common characteristic of impression evidence is that the pattern of interest is often latently hidden in structured clutter and severely occluded. This poses a major challenge to fully automated approaches for impression analysis. This talk discusses a generative approach that integrates the segmentation and classification of latent impressions in a common optimization framework.