Indonesian Doll Likeness
State of The Art LoRA (Low Rank Adaptation of Large Language Models) of Stable Diffusion model version 1.5 that has been trained on hundred image from Indonesian female photograph.
The current model is trained with 1.0e-6 learning rate, 10 epoch and equivalent of 2500 training steps.
Dataset has been preprocessed using Aspect Ratio Bucketing Tool.
Utilized BLIP2 as a part of the training process, as natural language prompts might be more effective.
This LoRA if combined with suitable model can generate Female Indonesian photorealistic images, with natural color tonality like photograph.
Model Details
- Trained and Maintained by: Gorilla Monsoon III
- Model type: Standard LoRA
Sample Images
Acknowledgement
- Kanbara (original creator of DollLikeness series)
- Stability AI, CompVis, RunwayML for Stable Diffusion ecosystems