erikpro007/OpenDatasetFlowersITI

---
license: apache-2.0
tags:
- Dataset
size_categories:
- 10K<n<100K
---

# 🌸 OpenDatasetFlowers – ImageToImage


A large-scale paired image dataset of AI-generated flower photographs and their corresponding line-art sketches, designed for image-to-image translation tasks such as sketch-to-photo synthesis, pix2pix training, and contour-guided generation.

## Dataset Summary

The dataset provided as `OpenDatasetFlowersITI.zip` contains approximately **80k images** (40k pairs) across **129 flower categories**. Each pair consists of a line-art sketch (INPUT) and the corresponding original colour photograph (OUTPUT). All images are uniform in format and resolution, ready to use without additional preprocessing.

| Property | Value |
|------------------|-------------------------------|
| Total Pairs | ~40,119 |
| Total Images | ~80,238 |
| Format | WebP |
| Resolution | 512 × 512 px |
| Categories | 129 |
| Pairs / Category | 311 |
| Archive Format | .zip |
| Size | ~3.8 GB |

## Structure

The dataset is split into two top-level folders. Files are matched by their identical path — `INPUT/Rose/0.webp` is the sketch counterpart of `OUTPUT/Rose/0.webp`.

```
OpenDatasetFlowersITI/
├── INPUT/ ← line-art sketches (white background)
│ ├── Rose/
│ │ ├── 0.webp
│ │ ├── 1.webp
│ │ └── ... 310.webp
│ ├── Tulip/
│ │ └── ...
│ └── ...
└── OUTPUT/ ← original colour photographs
├── Rose/
│ ├── 0.webp
│ ├── 1.webp
│ └── ... 310.webp
├── Tulip/
│ └── ...
└── ...
```

## Intended Use

This dataset is well-suited for training and evaluating:

- **Sketch-to-photo** translation models (pix2pix, ControlNet, etc.)
- **Contour-guided** image synthesis
- **Edge detection** and line-art extraction research
- **Domain adaptation** between photographic and artistic styles

## License

[Apache 2.0](https://opensource.org/licenses/Apache-2.0)