What is it about?
We combined well known k-same approach for preserving face privacy with the generative model which is able to generate new parametrized identities, where certain facial properties can be retained (e.g. expression of emotion). We evaluated the efficacy of our approach by performing tests with human observers as well as automated face recognition tools.
Featured Image
Why is it important?
In our work, we joined formal k-same anonymity scheme with the state of the art generative neural network. Qualitative results show almost no visual artefacts in the generated content and quantitative evaluation indicates that the recognition performance is indeed effectively reduced after deidentification process.
Perspectives
Generative neural networks are currently very popular amongst deep learning approaches. This paper displays the posibility of using such generative model for the purpose of preserving anonymity by producing joined identity from k-closest faces from closed set, that correspond to the original identity, thus enabling k-same deidentification scheme, while retaining non-identity related information (such as emotion).
Blaz Meden
Read the Original
This page is a summary of: Face Deidentification with Generative Deep Neural Networks , IET Signal Processing, May 2017, the Institution of Engineering and Technology (the IET),
DOI: 10.1049/iet-spr.2017.0049.
You can read the full text:
Contributors
The following have contributed to this page
