Divide and Conquer: a Two-Step Method for High Quality Face De-identification with Model Explainability.
Yunqian WenBo LiuJingyi CaoRong XieLi SongPublished in: ICCV (2023)
Keyphrases
- mathematical model
- high quality
- statistical model
- high order
- human head
- input data
- probabilistic model
- prior knowledge
- modeling method
- objective function
- cost function
- similarity measure
- study proposes
- network model
- evaluation method
- recognition algorithm
- theoretical analysis
- clustering method
- monte carlo simulation
- classification method
- evaluation model
- sensitivity analysis
- prior information
- markov model
- parameter estimation
- linear model
- iterative process
- prediction model
- regression analysis
- em algorithm
- highly accurate
- test data
- preprocessing
- dynamic programming
- support vector machine
- probability distribution
- estimation process
- bp neural network
- hybrid model
- detection method
- linear regression
- optimization method
- closed form
- kalman filter
- pairwise
- energy function