---
title: SynBag50K
canonical_url: "https://www.modelscope.cn/datasets/musinghead/SynBag50K"
md_url: "https://www.modelscope.cn/datasets/musinghead/SynBag50K.md"
repository: musinghead/SynBag50K
last_updated: 2026-10-07
license: "MIT License"
storage_size: "64 GB"
downloads: 333
stars: 0
---

# SynBag50K

> SynBag50K - musinghead 在 ModelScope 开源的数据集。SynBag50K is a synthetic dataset of 50,000 bag images generated using Z-Image-Turbo.

musinghead/SynBag50K 是 ModelScope 魔搭社区上的数据集，存储大小 64 GB，采用 MIT License 许可。

- **Repository**: musinghead/SynBag50K
- **License**: MIT License
- **Storage size**: 64 GB
- **Downloads**: 333
- **Stars**: 0
- **Last updated**: 2026-10-07

Source: https://www.modelscope.cn/datasets/musinghead/SynBag50K

---

# SynBag50K

SynBag50K is a synthetic dataset of 50,000 bag images generated using [Z-Image-Turbo](https://www.modelscope.cn/models/Tongyi-MAI/Z-Image-Turbo).

The dataset was used to train BagGEN ([Technical Report](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6259845), [Code](https://github.com/Musinghead/BagGEN), [Model](https://www.modelscope.cn/models/musinghead/BagGEN)), a fine-tuned IP-Adapter that conditions bag image generation on a texture patch, allowing textures and materials from reference images to guide the appearance of generated bags.

Each sample contains a generated bag image, a texture patch cropped from that image, structural conditions, and associated metadata, as illustrated below:

| Example | File Name | Description |
| --- | --- | --- |
| <img src="https://modelscope.cn/datasets/musinghead/SynBag50K/resolve/master/bag_item_sample/contour.png" alt="contour" width="200"> | contour.png | A contour condition derived from image.png |
| <img src="https://modelscope.cn/datasets/musinghead/SynBag50K/resolve/master/bag_item_sample/image.png" alt="image" width="200"> | image.png | A bag image generated using Z-Image-Turbo |
| <img src="https://modelscope.cn/datasets/musinghead/SynBag50K/resolve/master/bag_item_sample/mask.png" alt="mask" width="200"> | mask.png | A foreground mask for image.png |
| "left": 396, "upper": 490, <br> "right": 652, "lower": 746 | patch_coord.json | The crop coordinates of patch.png within image.png |
| <img src="https://modelscope.cn/datasets/musinghead/SynBag50K/resolve/master/bag_item_sample/patch.png" alt="patch" width="200"> | patch.png | A texture patch cropped from image.png |
| A nylon wide-base bucket bag <br> featuring tropical palm patterns, <br> detailed, centered, high-quality, <br> product photo, white background | prompt.txt | The text prompt used to generate image.png |
| "category": "wide-base bucket bag", <br> "decorative_theme": "tropical palm", <br> "material": "nylon" | property.json | Structured bag properties: category, decorative theme, and material |
| <img src="https://modelscope.cn/datasets/musinghead/SynBag50K/resolve/master/bag_item_sample/scribble.png" alt="scribble" width="200"> | scribble.png | A scribble condition derived from image.png |

## License and Data Provenance

SynBag50K is released under the [MIT License](LICENSE), covering the
dataset files and associated metadata provided in this repository.

Bag images were generated using Z-Image-Turbo. Foreground masks were
estimated using RMBG-2.0 and used in subsequent data preprocessing.
The models used during data preparation are not included in this
repository and remain subject to their respective licenses.