EXPERIMENTAL WORK
CURRENT
TFDNE was trained on publicly available datasets like Flickr portraits, which contains a disproportionately high number of Caucasian and Asian faces compared to other ethnicities. Ethnicities who are underrepresented also contain a high repetition of models used, further decreasing diversity within data sets. This bias causes the model to perpetuate existing societal inequalities, as it only creates variations based on the dominant demographics it was taught.
Facial merging code-based motion piece, detecting age and emotion and morphing images in a sequence together.
coded in Python ft. DeepFace, OpenCV, SciPy, NumPy, Dlib
(Post Processing within After Effects for use in a game)
Video by me, decaying + signs spawning on video, coded in python
Compiled in Touchdesigner, videos sourced from me + stock sites.
Testing... BNO055 Arduino IDE Unreal Engine