DeeplearningDeepfakeComputer Vision
Preserving the Evidence: Artifact-Aware Preprocessing for Robust Deepfake Detection
This study proves that intelligent preprocessing, not just model architecture, is the key to robust deepfake detection. Using an adaptive wavelet-based pipeline, we achieved 99.96% AUROC and perfect 100% precision at the video level on the Celeb-DF v2 dataset.
Andrew Alfonso Lie2 min read
