(CASSI)
CASSI runs on a neural network, a fine-tuned version of the ResNet50 CNN, trained with 2000-4000 images of synthetic faces as well as celebrity face datasets.
The model is able to identify faces I would deem attractive to an 80-90% accuracy. With a built in camera, live video uses OpenCV to detect faces and send them to the model for judgement, indicating an answer with CASSI’s light-up face.
3D printed PVA, synthetic hair, cardboard, scar tape, Jetsen Nano with RP Camera
Data from thispersondoesnotexist.com and Kaustubh Dhote’s face dataset
Python & TensorFlow and PyTorch
Data from thispersondoesnotexist.com and Kaustubh Dhote’s face dataset
Python & TensorFlow and PyTorch
AND
DEVELOPMENT
FULL DESIGN BRIEF: IDEATION, RESEARCH, DESIGN, DEVELOPMENT
PROTOTYPING