jaddai/openbrush-landscapes

---
license: mit
task_categories:
- image-to-text
- text-to-image
- image-classification
language:
- en
tags:
- art
- fine-art
- paintings
- public-domain
- landscape
- scenery
- nature
- genre
- structured-captions
- image-captioning
- diffusion
- vlm
size_categories:
- 10K<n<100K
configs:
- config_name: default
data_files:
- split: train
path: data/train-*.parquet
---

# OpenBrush Landscapes

Every landscape painting from OpenBrush-75K — across all artists, movements, and centuries. Largest single-genre subset.

> **Curated subset of [jaddai/openbrush-75k](https://huggingface.co/datasets/jaddai/openbrush-75k).** Same MIT license, same caption schema, same VLM (Qwen3-VL-30B-A3B). This subset exists so you don't have to download 75,313 images to get to the 12,612 you actually want.

## Why this subset

Every landscape across the parent dataset's full range — Romantic wildernesses, Impressionist plein-air, Hudson River School-era American vistas, Northern European countryside, Mediterranean coast, Russian winter, Japanese woodblock landscape. Evergreen LoRA training target.

**Filter equivalent on parent:** `genre == "landscape"`

## Statistics

- **Total images:** 12,612
- **Source:** OpenBrush-75K v1.1 (filtered)
- **License:** MIT
- **Caption schema:** v2 structured (TAGS / SUBJECT / ACTION / SETTING / MOOD / STYLE / LIGHTING / COLOR / COMPOSITION)

## Caption schema

Same as the parent. Each row has `caption_full` plus 9 individual fields:

```
**TAGS:** comma-separated standardized descriptive tags
**SUBJECT:** Description of the main subject
**ACTION:** Movement, gesture, what is happening
**SETTING:** Environment and context
**MOOD:** Emotional tone and atmosphere
**STYLE:** Artistic technique and stylistic analysis
**LIGHTING:** Direction, quality, temperature
**COLOR:** Palette and color relationships
**COMPOSITION:** Framing, balance, focal points
```

## Schema

| Field | Type | Description |
|-------|------|-------------|
| `id` | string | Unique identifier |
| `image` | Image | The artwork (PNG bytes) |
| `width` | int | Image width in pixels |
| `height` | int | Image height in pixels |
| `artist` | string | Artist name |
| `style` | string | Art movement/style |
| `genre` | string | Genre |
| `tags` | list[string] | Standardized descriptive tags |
| `subject`, `action`, `setting`, `mood`, `style_description`, `lighting`, `color`, `composition` | string | Per-aspect caption fields |
| `caption_full` | string | Full structured caption |
| `source_file` | string | Original filename |

## Loading

```python
from datasets import load_dataset

dataset = load_dataset("jaddai/openbrush-landscapes")
print(len(dataset["train"])) # 12,612

example = dataset["train"][0]
print(example["caption_full"])
```

## Relationship to the parent dataset

This is a filtered slice of [jaddai/openbrush-75k](https://huggingface.co/datasets/jaddai/openbrush-75k) v1.1. All images, captions, and metadata are identical to those in the parent — no recaptioning or modification.

If you want the **complete** OpenBrush corpus (75,313 paintings), use the parent. If you want **just this slice**, use this subset to save bandwidth and disk.

### Sibling subsets

- [openbrush-van-gogh](https://huggingface.co/datasets/jaddai/openbrush-van-gogh)
- [openbrush-monet](https://huggingface.co/datasets/jaddai/openbrush-monet)
- [openbrush-renoir](https://huggingface.co/datasets/jaddai/openbrush-renoir)
- [openbrush-rembrandt](https://huggingface.co/datasets/jaddai/openbrush-rembrandt)
- [openbrush-impressionism](https://huggingface.co/datasets/jaddai/openbrush-impressionism)
- [openbrush-renaissance](https://huggingface.co/datasets/jaddai/openbrush-renaissance)
- [openbrush-ukiyo-e](https://huggingface.co/datasets/jaddai/openbrush-ukiyo-e)
- [openbrush-baroque](https://huggingface.co/datasets/jaddai/openbrush-baroque)
- [openbrush-landscapes](https://huggingface.co/datasets/jaddai/openbrush-landscapes)
- [openbrush-portraits](https://huggingface.co/datasets/jaddai/openbrush-portraits)
- [openbrush-religious-art](https://huggingface.co/datasets/jaddai/openbrush-religious-art)
- [openbrush-anonymous-masters](https://huggingface.co/datasets/jaddai/openbrush-anonymous-masters)
- [openbrush-impressionist-landscapes](https://huggingface.co/datasets/jaddai/openbrush-impressionist-landscapes)

## License

MIT. Underlying artworks are public domain (all artists died before 1954). Structured captions are original works released under MIT.

## Citation

```bibtex
@dataset{openbrush_openbrush_landscapes,
title={OpenBrush Landscapes},
author={jaddai},
year={2026},
publisher={Hugging Face},
url={https://huggingface.co/datasets/jaddai/openbrush-landscapes},
note={Curated subset of openbrush-75k}
}
```

## Acknowledgments

- Original artwork images derived from public domain art collections
- Captions generated using Qwen3-VL-30B-A3B
- Tag standardization assisted by LLM-based semantic mapping