README.md

---
datasets:
- cQueenccc/Vivian-Blip-Captions
language:
- en
pipeline_tag: text-to-image
---

# Disclaimer
This was inspired from https://github.com/YaYaB/finetune-diffusion

# Model Card for Finetuning Stable Diffusion on Vivian Maier's photographs
The main goal is to fine-tune the Stable Diffusion model to generate images reflecting the distinct photographic style of Vivian Maier.

And I chose to utilize a Jupyter Notebook to make the fine-tuning process accessible and easy to understand, particularly for those new to the diffusion pipeline and hugging face API.

# Requirements
To launch the finetuning with a batch_size of 1 you need to have a gpu with at least 24G VRAM (you can use accumulating gradient to simulate higher batch size)

Make sure that you have enough disk space, the model uses ~11Gb

## Examples(at epoch 90)

![vv1.jpg](https://huggingface.co/cQueenccc/Fine-Tune-Diffusion-Vivian/resolve/main/eval/A%20woman%20walking%20down%20the%20street/A%20woman%20walking%20down%20the%20street_90_000000.png)
> A woman walking down a street

![vv2.jpg](https://huggingface.co/cQueenccc/Fine-Tune-Diffusion-Vivian/resolve/main/eval/a%20group%20of%20people%20getting%20on%20a%20bus/a%20group%20of%20people%20getting%20on%20a%20bus_90_000000.png)
> a group of people getting on a bus

![vv3.jpg](https://huggingface.co/cQueenccc/Fine-Tune-Diffusion-Vivian/resolve/main/eval/two%20men%20working%20on%20a%20construction%20site/two%20men%20working%20on%20a%20construction%20site_90_000000.png)
> two man working on a constructing site

## Citation

If you use this dataset, please cite it as:

```
@misc{cqueenccc2023vivian,
      author = {cQueenccc},
      title = {Finetuning Stable Diffusion on Vivian Maier's photographs},
      year={2023},
      howpublished= {\url{https://huggingface.co/cQueenccc/Fine-Tune-Diffusion-Vivian/}}
} 
```