Artificial Fingerprinting for Generative Models: Rooting Deepfake Attribution in Training Data.
Ning YuVladislav SkripniukSahar AbdelnabiMario FritzPublished in: ICCV (2021)
Keyphrases
- generative model
- training data
- prior knowledge
- discriminative methods
- probabilistic model
- mixture model
- semi supervised
- learning algorithm
- discriminative learning
- em algorithm
- discriminative models
- supervised learning
- data sets
- training set
- labeled data
- generative and discriminative models
- learned models
- maximum entropy principle
- hierarchical hidden markov models
- conditional random fields
- training examples
- unlabeled data
- classification accuracy
- naive bayes
- topic models
- training samples
- semi supervised learning
- information extraction
- decision trees
- class labels
- learning tasks
- object categories
- multiscale
- object parts
- representational power
- active learning
- naive bayes models