---
title: InSubject-Dataset
canonical_url: "https://www.modelscope.cn/datasets/AI-ModelScope/InSubject-Dataset"
md_url: "https://www.modelscope.cn/datasets/AI-ModelScope/InSubject-Dataset.md"
repository: AI-ModelScope/InSubject-Dataset
last_updated: 2025-10-22
license: cc-by-4.0
storage_size: "259 MB"
downloads: 137
stars: 1
---

# InSubject-Dataset

> InSubject-Dataset - AI-ModelScope 在 ModelScope 开源的数据集。This dataset contains 1638 approved image generation samples with detailed metadata.

AI-ModelScope/InSubject-Dataset 是 ModelScope 魔搭社区上的数据集，存储大小 259 MB，采用 cc-by-4.0 许可。

- **Repository**: AI-ModelScope/InSubject-Dataset
- **License**: cc-by-4.0
- **Storage size**: 259 MB
- **Downloads**: 137
- **Stars**: 1
- **Last updated**: 2025-10-22

Source: https://www.modelscope.cn/datasets/AI-ModelScope/InSubject-Dataset

---

# InSubject-Dataset

## Dataset Description

This dataset contains 1638 approved image generation samples with detailed metadata.

### Dataset Summary

- **Total Images**: 1638
- **Source Datasets**: 2, 3, 14
- **Review Status**: All items are approved/accepted

### Fields

- `image`: The generated image
- `filename`: Original filename
- `prompt`: Generation prompt
- `style_reference`: Style reference/tag
- `character_reference`: Character reference/tag
- `scene_reference`: Scene reference/tag
- `width`: Image width in pixels
- `height`: Image height in pixels
- `orientation`: Image orientation (portrait/landscape/square)
- `size_category`: Size category
- `based_on`: Reference image filename (if applicable)
- `generation_prompt`: Transformed generation prompt
- `generation_reasoning`: Generation reasoning/notes
- `dataset_id`: Source dataset ID

## Usage

```python
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("peteromallet/InSubject-Dataset")

# Access an item
item = dataset['train'][0]
image = item['image']
prompt = item['prompt']
style = item['style_reference']
```

## Citation

If you use this dataset, please cite appropriately and follow the license terms.