- Nick Hale (nickhale@jhu.edu)
- David Li (dzli@jhu.edu)
- Aurik Sarker (asarker1@jhu.edu)
We investigated the effectiveness of Neural Style Transfer for transferring the style of photographs to artwork to improve object detection in abstract artwork. This attempts to solve the cross-depiction problem, the problem of recognizing objects in images independent of what style (drawn, photographed, painted) they are shown in.
We tested various methods of style removal and compared their performances on YOLO (You Only Look Once), a state-of-the-art object detection model.
Our final paper can be found here.