IST - Style Transfer with Instance Segmentation
- Resource Type
- Conference
- Authors
- Stahl, Fabian; Meyer, Melina; Schwanecke, Ulrich
- Source
- 2019 11th International Symposium on Image and Signal Processing and Analysis (ISPA) Image and Signal Processing and Analysis (ISPA), 2019 11th International Symposium on. :277-281 Sep, 2019
- Subject
- Signal Processing and Analysis
Image segmentation
Training
Proposals
Real-time systems
Feature extraction
Computer vision
Signal processing
stylization
segmentation
CNN
- Language
- ISSN
- 1849-2266
We present a technique to highlight people on images in different styles. Our instance style transfer model (IST) combines an instance segmentation model and a stylization model. Image parts that contain people are found by a segmentation network and then stylized by a style transfer network. The modified image parts are combined with the original image. The resulting image contains people highlighted in different ways, while the background remains unchanged.