[arXiv] [BibTeX] [Gitlab] [Demo]
| State file | Size | Where to place | Download |
|---|---|---|---|
| stylematte_pure.pth | 133.6 MB | stylematte/checkpoints/ |
(https://github.com/chroneus/stylematte/releases/download/weights/stylematte_pure.pth) |
| stylematte_synth.pth | 133.6 MB | stylematte/checkpoints/ |
(https://github.com/chroneus/stylematte/releases/download/weights/stylematte_synth.pth) |
| animals.pkl | 300.5 MB | stylegan3 |
(https://github.com/chroneus/stylematte/releases/download/weights/animals.pkl) |
| humans.pkl | 281.1 MB | stylegan3 |
(https://github.com/chroneus/stylematte/releases/download/weights/humans.pkl) |
To synthesize synthetic dataset of RGBA images, move to stylegan3 and run synthesize.py. You should create conda environment from stylegan3/environment.yml. You can also generate images with different truncation and seed values using gen_images.py .
conda activate stylegan3
cd stylegan3
python synthesize.pyTo change image background on synthetic dataset, run
python visualizer.pyIn the GUI you can choose model weights, background picture, truncation value and other visualization parameters.
To test our model, change directory to stylematte and run test.py. You can modify test.yaml file for your datasets and models.
cd stylematte
python test.pyThe report directory is stylematte/report/. See report examples there.
This work is licensed under a variant of Creative Commons Attribution-ShareAlike 4.0 International License.
Please see the specific license.