no code implementations • 19 Apr 2023 • Rosaura G. VidalMata, Priscila Saboia, Daniel Moreira, Grant Jensen, Jason Schlessman, Walter J. Scheirer
To this end, we introduce an anomaly enhancement loss that, when used with a residual architecture, improves the performance of different detection algorithms with a minimal introduction of false positives on the non-manipulated data.
no code implementations • 3 Apr 2023 • Jacob Piland, Christopher Sweet, Priscila Saboia, Charles Vardeman II, Adam Czajka
Energy-based models (EBM) have become increasingly popular within computer vision.