rinongal/StyleGAN-nada
Text-guided domain adaptation method that shifts pre-trained StyleGAN generators to new visual domains using CLIP without requiring any training images.
Not currently ranked — collecting fresh signals.
star history
StyleGAN-NADA adapts image generators to new domains using only natural language text prompts and CLIP guidance. The method leverages the semantic power of CLIP models to shift StyleGAN to diverse domains characterized by different styles and shapes, without collecting any images from the target domain. This zero-shot domain adaptation approach enables practitioners to repurpose pre-trained generators through natural language descriptions alone.
Frequently asked
- What is rinongal/StyleGAN-nada?
- Text-guided domain adaptation method that shifts pre-trained StyleGAN generators to new visual domains using CLIP without requiring any training images.
- Is StyleGAN-nada open source?
- Yes — rinongal/StyleGAN-nada is open source, released under the MIT license.
- What language is StyleGAN-nada written in?
- rinongal/StyleGAN-nada is primarily written in Python.
- How popular is StyleGAN-nada?
- rinongal/StyleGAN-nada has 1.2k stars on GitHub.
- Where can I find StyleGAN-nada?
- rinongal/StyleGAN-nada is on GitHub at https://github.com/rinongal/StyleGAN-nada.