PixArt-alpha / PixArt-sigma

PixArt-Σ: Weak-to-Strong Training of Diffusion Transformer for 4K Text-to-Image Generation
https://pixart-alpha.github.io/PixArt-sigma-project/
GNU Affero General Public License v3.0
1.44k stars 67 forks source link

If I want to add a new person in the generate model, how many data should I prepare for finetuning ? #107

Open ApolloRay opened 1 month ago

ApolloRay commented 1 month ago

Now, if I input obama or trump, it will draw the right person. If I want to add some new person, how many data should I prepare ?

lawrence-cj commented 1 week ago

I'm not sure how to only "add" a new person into the model. Maybe a dreambooth-related method is more reasonable? If you fine-tune only on the specific dataset, the model's capability maybe degrated.