Currently still a work in progress
Currently getting varying results when using just a single training image. Most manga characters arent as clean as the ones demonstrated in the paper. Works best if target Image has little to no screentones.
Paper: https://arxiv.org/pdf/1706.06918.pdf
TODO:
- ScreenTone Removal
- Reapplying Screentones/Shading
- Fix Trapped Ball Segmentation
- Test on simpler Datasets
