Social Data Analyzer

The Goal of this Paper is to implement a tool to analyze social datasets containing videos/images of faces. - Diversity (Age/Gender/Ethnicity/…) - Optional: Leakage (Face recognition between test, validation, and training data). Tasks for this Thesis:

1. Research existing methods to examine (social) datasets for diversity
   • Choose one or more suitable model(s)
2. Research existing social datasets (including VVAD-LRS3 and WildVVAD)
3. Implement a generic python library to analyze facial image and video data
   • Define API
   • Define possible input output formats
   • CLI (optional GUI)
4. Use the library to compare VVAD-LRS3 and WildVVAD (among others)

Staring Points:

• FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age for Bias Measurement and Mitigation

This could be published in collaboration with the AG Diversity