This project aims to construct a facial image synthetic data engine in Unity (FaceX) similar to existing PersonX and VehicleX datasets to be usable for person Re-ID, emotion recognition, age/gender recognition etc.
For developing synthetic face emotion dataset, writing a literature survey is the first step to know more detail about this project, and then the preliminary list of steps are as following:
-
learn Unity, and find face models / tools to construct face models (eg naturalfront.com);
-
construct sufficient face identities with variations and generate a large synthetic data set;
-
implement an architecture from literature to compare results of synthetic and real training.
-
Develop an approach with appropriate trade-offs of properties.
If time permits use some tools to include functionality to modify face adornments (hats, glasses etc).
Finally, delivering a report.
https://anu.zoom.us/j/603332221?pwd=QWJ4d1FySXVSSXEraWhoN3B3NTR4Zz09