Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -4,54 +4,3 @@ language:
|
|
| 4 |
- en
|
| 5 |
pretty_name: e
|
| 6 |
---
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
# Dataset Card: MangaDF Text-to-Image Prompts Dataset
|
| 10 |
-
|
| 11 |
-
## Dataset Description
|
| 12 |
-
|
| 13 |
-
**Image Source:** Images generated by [alvdansen/BandW-Manga](https://huggingface.co/alvdansen/BandW-Manga) weights on [ChanY/Stable-Flash-Lighting](https://huggingface.co/ChanY/Stable-Flash-Lighting) model
|
| 14 |
-
**Prompt Source:** ChatGPT
|
| 15 |
-
|
| 16 |
-
## Overview
|
| 17 |
-
|
| 18 |
-
The MangaDF Text-to-Image Prompts Dataset is a collection of text prompts paired with corresponding images. The images in this dataset were generated using the [alvdansen/BandW-Manga](https://huggingface.co/alvdansen/BandW-Manga) weights applied to the [ChanY/Stable-Flash-Lighting](https://huggingface.co/ChanY/Stable-Flash-Lighting) diffusion model. This dataset is designed to aid in training and evaluating models for text-to-image generation tasks, specifically focusing on bold line portrait illustrations in a manga style.
|
| 19 |
-
|
| 20 |
-
## Dataset Structure
|
| 21 |
-
|
| 22 |
-
### Data Fields
|
| 23 |
-
|
| 24 |
-
- **prompt:** A string containing the text prompt used to generate the image.
|
| 25 |
-
- **image:** The generated image corresponding to the text prompt, stored as a path to the image file.
|
| 26 |
-
|
| 27 |
-
### Example
|
| 28 |
-
|
| 29 |
-
Here is an example of a data entry in this dataset:
|
| 30 |
-
|
| 31 |
-
```python
|
| 32 |
-
from datasets import load_dataset
|
| 33 |
-
dataset = load_dataset("Chan-Y/Mangaset", split="train")
|
| 34 |
-
|
| 35 |
-
from PIL import Image
|
| 36 |
-
import io
|
| 37 |
-
|
| 38 |
-
image_bytes = dataset["image"][3]
|
| 39 |
-
image = Image.open(io.BytesIO(image_bytes))
|
| 40 |
-
image
|
| 41 |
-
```
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
## Creation Process
|
| 45 |
-
|
| 46 |
-
1. **Model Selection:** The `alvdansen/BandW-Manga` weights were applied to the `ChanY/Stable-Flash-Lighting` diffusion model.
|
| 47 |
-
2. **Image Generation:** Images were generated using a variety of text prompts, focusing on producing high-quality manga-style bold line portraits.
|
| 48 |
-
3. **Dataset Compilation:** Each generated image was paired with its corresponding text prompt, and the pairs were compiled into a Parquet file for efficient storage and access.
|
| 49 |
-
|
| 50 |
-
## Acknowledgements
|
| 51 |
-
|
| 52 |
-
- Image generation weights: [alvdansen/BandW-Manga](https://huggingface.co/alvdansen/BandW-Manga)
|
| 53 |
-
- Diffusion model: [ChanY/Stable-Flash-Lighting](https://huggingface.co/ChanY/Stable-Flash-Lighting)
|
| 54 |
-
|
| 55 |
-
## Contact
|
| 56 |
-
|
| 57 |
-
For any questions or issues, please contact [me](https://www.linkedin.com/in/chanyalcin/)
|
|
|
|
| 4 |
- en
|
| 5 |
pretty_name: e
|
| 6 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|