Unsupervised privacy-enhancement of face representations using similarity-sensitive noise transformations.
Philipp TerhörstNaser DamerFlorian KirchbuchnerArjan KuijperPublished in: Appl. Intell. (2019)
Keyphrases
- word sense induction
- similarity measure
- protect sensitive
- pointwise mutual information
- image transformations
- structural similarity
- noise level
- unsupervised learning
- privacy preserving
- invariant representations
- security issues
- distance measure
- signal to noise ratio
- sensitive information
- semi supervised
- supervised learning
- missing data
- image enhancement
- facial images
- social networks
- distance function
- facial expressions
- private data
- euclidean distance
- gaussian noise
- human faces
- semantic similarity
- noise reduction
- noisy data
- face images
- noisy environments
- privacy preservation
- keypoints
- shape similarity
- unsupervised manner