# AnyInsertion_V1

**Repository Path**: hf-datasets/AnyInsertion_V1

## Basic Information

- **Project Name**: AnyInsertion_V1
- **Description**: Mirror of https://huggingface.co/datasets/WensongSong/AnyInsertion_V1
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-07-15
- **Last Updated**: 2025-07-15

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

---
dataset_info:
- config_name: mask_prompt
  features:
  - name: id
    dtype: string
  - name: split
    dtype: string
  - name: category
    dtype: string
  - name: label
    dtype: string
  - name: ref_image
    dtype: image
  - name: ref_mask
    dtype: image
  - name: tar_image
    dtype: image
  - name: tar_mask
    dtype: image
  splits:
  - name: train
    num_bytes: 58055297653.968
    num_examples: 58318
  download_size: 57605566640
  dataset_size: 58055297653.968
- config_name: text_prompt
  features:
  - name: id
    dtype: string
  - name: split
    dtype: string
  - name: category
    dtype: string
  - name: label
    dtype: string
  - name: src_label
    dtype: string
  - name: src_type
    dtype: string
  - name: ref_image
    dtype: image
  - name: ref_mask
    dtype: image
  - name: src_image
    dtype: image
  - name: tar_image
    dtype: image
  splits:
  - name: train
    num_bytes: 183941024225.888
    num_examples: 78234
  download_size: 183594500291
  dataset_size: 183941024225.888
configs:
- config_name: mask_prompt
  data_files:
  - split: train
    path: mask_prompt/train-*
- config_name: text_prompt
  data_files:
  - split: train
    path: text_prompt/train-*
---



---
license: mit
task_categories:
- image-to-image
language:
- en
pretty_name: a
size_categories:
- 10M<n<100M
---
# AnyInsertion
<p align="center">
<a href="https://song-wensong.github.io/"><strong>Wensong Song</strong></a>
·
<a href="https://openreview.net/profile?id=~Hong_Jiang4"><strong>Hong Jiang</strong></a>
·
<a href="https://z-x-yang.github.io/"><strong>Zongxing Yang</strong></a>
·
<a href="https://scholar.google.com/citations?user=WKLRPsAAAAAJ&hl=en"><strong>Ruijie Quan</strong></a>
·
<a href="https://scholar.google.com/citations?user=RMSuNFwAAAAJ&hl=en"><strong>Yi Yang</strong></a>
<br>
<br>
<a href="https://arxiv.org/pdf/2504.15009" style="display: inline-block; margin-right: 10px;">
  <img src='https://img.shields.io/badge/arXiv-InsertAnything-red?color=%23aa1a1a' alt='Paper PDF'>
</a>
<a href='https://song-wensong.github.io/insert-anything/' style="display: inline-block; margin-right: 10px;">
  <img src='https://img.shields.io/badge/Project%20Page-InsertAnything-cyan?logoColor=%23FFD21E&color=%23cbe6f2' alt='Project Page'>
</a>
<a href='https://github.com/song-wensong/insert-anything' style="display: inline-block;">
  <img src='https://img.shields.io/badge/GitHub-InsertAnything-black?logoColor=23FFD21E&color=%231d2125'>
</a>
<br>
<b>Zhejiang University &nbsp; | &nbsp; Harvard University &nbsp; | &nbsp;  Nanyang Technological University </b>
</p>

## News
* <span style="color:red">**[2025.5.9]** Release new **AnyInsertion** v1 text- and mask-prompt dataset on [HuggingFace](https://huggingface.co/datasets/WensongSong/AnyInsertion_V1).</span>
* **[2025.5.7]** Release **AnyInsertion** v1 text prompt dataset on HuggingFace.
* **[2025.4.24]** Release **AnyInsertion** v1 mask prompt dataset on HuggingFace.


## Summary
This is the dataset proposed in our paper [**Insert Anything: Image Insertion via In-Context Editing in DiT**](https://arxiv.org/abs/2504.15009)

AnyInsertion dataset consists of training and testing subsets. The training set includes 136,385 samples across two prompt types: 58,188 mask-prompt image pairs and 78,197 text-prompt image pairs;the test set includes 158 data pairs: 120 mask-prompt pairs and 38 text-prompt pairs.

AnyInsertion dataset covers diverse categories including human subjects, daily necessities, garments, furniture, and various objects.


![alt text](dataset_categories.png)


## Directory

```




├── text_prompt/
│   ├── train/
│   │   ├── accessory/
│   │   │   ├── ref_image/      # Reference image containing the element to be inserted
│   │   │   ├── ref_mask/       # The mask corresponding to the inserted element
│   │   │   ├── tar_image/      # Ground truth
│   │   │   └── src_image/      # Source images
│   │   │       ├── add/        # Source image with the inserted element from Ground Truth removed
│   │   │       └── replace/    # Source image where the inserted element in Ground Truth is replaced
│   │   ├── object/
│   │   │   ├── ref_image/      
│   │   │   ├── ref_mask/      
│   │   │   ├── tar_image/      
│   │   │   └── src_image/      
│   │   │       ├── add/       
│   │   │       └── replace/    
│   │   └── person/
│   │       ├── ref_image/     
│   │       ├── ref_mask/       
│   │       ├── tar_image/     
│   │       └── src_image/    
│   │           ├── add/       
│   │           └── replace/  
│   └── test/
│       ├── garment/
│       │   ├── ref_image/      
│       │   ├── ref_mask/      
│       │   ├── tar_image/      
│       │   └── src_image/      
│       └── object/
│           ├── ref_image/     
│           ├── ref_mask/       
│           ├── tar_image/     
│           └── src_image/      
│
├── mask_prompt/
│   ├── train/
│   │   ├── accessory/
│   │   │   ├── ref_image/     
│   │   │   ├── ref_mask/      
│   │   │   ├── tar_image/     
│   │   │   ├── tar_mask/      # The mask corresponding to the edited area of target image
│   │   ├── object/
│   │   │   ├── ref_image/    
│   │   │   ├── ref_mask/      
│   │   │   ├── tar_image/     
│   │   │   ├── tar_mask/      
│   │   └── person/
│   │       ├── ref_image/     
│   │       ├── ref_mask/     
│   │       ├── tar_image/    
│   │       ├── tar_mask/   
│   └── test/
│       ├── garment/
│       │   ├── ref_image/   
│       │   ├── ref_mask/    
│       │   ├── tar_image/    
│       │   ├── tar_mask/    
│       ├── object/
│       │   ├── ref_image/    
│       │   ├── ref_mask/    
│       │   ├── tar_image/     
│       │   ├── tar_mask/    
│       └── person/
│           ├── ref_image/   
│           ├── ref_mask/   
│           ├── tar_image/  
│           ├── tar_mask/    

    




```



## Example

<div style="display: flex; text-align: center; align-items: center; justify-content: space-between;">
  <figure style="margin: 10px; width: calc(25% - 20px);">
    <img src="examples/ref_image.png" alt="Ref_image" style="width: 100%;">
    <figcaption>Ref_image</figcaption>
  </figure>
  <figure style="margin: 10px; width: calc(25% - 20px);">
    <img src="examples/ref_mask.png" alt="Ref_mask" style="width: 100%;">
    <figcaption>Ref_mask</figcaption>
  </figure>
  <figure style="margin: 10px; width: calc(25% - 20px);">
    <img src="examples/tar_image.png" alt="Tar_image" style="width: 100%;">
    <figcaption>Tar_image</figcaption>
  </figure>
  <figure style="margin: 10px; width: calc(25% - 20px);">
    <img src="examples/tar_mask.png" alt="Tar_mask" style="width: 100%;">
    <figcaption>Tar_mask</figcaption>
  </figure>
  <figure style="margin: 10px; width: calc(25% - 20px);">
    <img src="examples/add.png" alt="Add" style="width: 100%;">
    <figcaption>Add</figcaption>
  </figure>
  <figure style="margin: 10px; width: calc(25% - 20px);">
    <img src="examples/replace.png" alt="Replace" style="width: 100%;">
    <figcaption>Replace</figcaption>
  </figure>
</div>

### Text Prompt
Add Prompt: Add [label from `tar_image` (in label.json) ]</p>
Replace Prompt: Replace [label from `src_image` (in src_image/replace/replace_label.json) ] with [label from `tar_image` (in label.json) ]


## Citation
```
@article{song2025insert,
  title={Insert Anything: Image Insertion via In-Context Editing in DiT},
  author={Song, Wensong and Jiang, Hong and Yang, Zongxing and Quan, Ruijie and Yang, Yi},
  journal={arXiv preprint arXiv:2504.15009},
  year={2025}
}
```