WensongSong commited on
Commit
f4ce60c
·
verified ·
1 Parent(s): 8011cf3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +184 -0
README.md CHANGED
@@ -62,3 +62,187 @@ configs:
62
  - split: train
63
  path: text_prompt/train-*
64
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
62
  - split: train
63
  path: text_prompt/train-*
64
  ---
65
+
66
+
67
+
68
+ ---
69
+ license: mit
70
+ task_categories:
71
+ - image-to-image
72
+ language:
73
+ - en
74
+ pretty_name: a
75
+ size_categories:
76
+ - 10M<n<100M
77
+ ---
78
+ # AnyInsertion
79
+ <p align="center">
80
+ <a href="https://song-wensong.github.io/"><strong>Wensong Song</strong></a>
81
+ ·
82
+ <a href="https://openreview.net/profile?id=~Hong_Jiang4"><strong>Hong Jiang</strong></a>
83
+ ·
84
+ <a href="https://z-x-yang.github.io/"><strong>Zongxing Yang</strong></a>
85
+ ·
86
+ <a href="https://scholar.google.com/citations?user=WKLRPsAAAAAJ&hl=en"><strong>Ruijie Quan</strong></a>
87
+ ·
88
+ <a href="https://scholar.google.com/citations?user=RMSuNFwAAAAJ&hl=en"><strong>Yi Yang</strong></a>
89
+ <br>
90
+ <br>
91
+ <a href="https://arxiv.org/pdf/2504.15009" style="display: inline-block; margin-right: 10px;">
92
+ <img src='https://img.shields.io/badge/arXiv-InsertAnything-red?color=%23aa1a1a' alt='Paper PDF'>
93
+ </a>
94
+ <a href='https://song-wensong.github.io/insert-anything/' style="display: inline-block; margin-right: 10px;">
95
+ <img src='https://img.shields.io/badge/Project%20Page-InsertAnything-cyan?logoColor=%23FFD21E&color=%23cbe6f2' alt='Project Page'>
96
+ </a>
97
+ <a href='https://github.com/song-wensong/insert-anything' style="display: inline-block;">
98
+ <img src='https://img.shields.io/badge/GitHub-InsertAnything-black?logoColor=23FFD21E&color=%231d2125'>
99
+ </a>
100
+ <br>
101
+ <b>Zhejiang University &nbsp; | &nbsp; Harvard University &nbsp; | &nbsp; Nanyang Technological University </b>
102
+ </p>
103
+
104
+ ## News
105
+ * <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>
106
+ * **[2025.5.7]** Release **AnyInsertion** v1 text prompt dataset on HuggingFace.
107
+ * **[2025.4.24]** Release **AnyInsertion** v1 mask prompt dataset on HuggingFace.
108
+
109
+
110
+ ## Summary
111
+ This is the dataset proposed in our paper [**Insert Anything: Image Insertion via In-Context Editing in DiT**](https://arxiv.org/abs/2504.15009)
112
+
113
+ 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.
114
+
115
+ AnyInsertion dataset covers diverse categories including human subjects, daily necessities, garments, furniture, and various objects.
116
+
117
+
118
+ ![alt text](dataset_categories.png)
119
+
120
+
121
+ ## Directory
122
+
123
+ ```
124
+
125
+
126
+
127
+
128
+ ├── text_prompt/
129
+ │ ├── train/
130
+ │ │ ├── accessory/
131
+ │ │ │ ├── ref_image/ # Reference image containing the element to be inserted
132
+ │ │ │ ├── ref_mask/ # The mask corresponding to the inserted element
133
+ │ │ │ ├── tar_image/ # Ground truth
134
+ │ │ │ └── src_image/ # Source images
135
+ │ │ │ ├── add/ # Source image with the inserted element from Ground Truth removed
136
+ │ │ │ └── replace/ # Source image where the inserted element in Ground Truth is replaced
137
+ │ │ ├── object/
138
+ │ │ │ ├── ref_image/
139
+ │ │ │ ├── ref_mask/
140
+ │ │ │ ├── tar_image/
141
+ │ │ │ └── src_image/
142
+ │ │ │ ├── add/
143
+ │ │ │ └── replace/
144
+ │ │ └── person/
145
+ │ │ ├── ref_image/
146
+ │ │ ├── ref_mask/
147
+ │ │ ├── tar_image/
148
+ │ │ └── src_image/
149
+ │ │ ├── add/
150
+ │ │ └── replace/
151
+ │ └── test/
152
+ │ ├── garment/
153
+ │ │ ├── ref_image/
154
+ │ │ ├── ref_mask/
155
+ │ │ ├── tar_image/
156
+ │ │ └── src_image/
157
+ │ └── object/
158
+ │ ├── ref_image/
159
+ │ ├── ref_mask/
160
+ │ ├── tar_image/
161
+ │ └── src_image/
162
+
163
+ ├── mask_prompt/
164
+ │ ├── train/
165
+ │ │ ├── accessory/
166
+ │ │ │ ├── ref_image/
167
+ │ │ │ ├── ref_mask/
168
+ │ │ │ ├── tar_image/
169
+ │ │ │ ├── tar_mask/ # The mask corresponding to the edited area of target image
170
+ │ │ ├── object/
171
+ │ │ │ ├── ref_image/
172
+ │ │ │ ├── ref_mask/
173
+ │ │ │ ├── tar_image/
174
+ │ │ │ ├── tar_mask/
175
+ │ │ └── person/
176
+ │ │ ├── ref_image/
177
+ │ │ ├── ref_mask/
178
+ │ │ ├── tar_image/
179
+ │ │ ├── tar_mask/
180
+ │ └── test/
181
+ │ ├── garment/
182
+ │ │ ├── ref_image/
183
+ │ │ ├── ref_mask/
184
+ │ │ ├── tar_image/
185
+ │ │ ├── tar_mask/
186
+ │ ├── object/
187
+ │ │ ├── ref_image/
188
+ │ │ ├── ref_mask/
189
+ │ │ ├── tar_image/
190
+ │ │ ├── tar_mask/
191
+ │ └── person/
192
+ │ ├── ref_image/
193
+ │ ├── ref_mask/
194
+ │ ├── tar_image/
195
+ │ ├── tar_mask/
196
+
197
+
198
+
199
+
200
+
201
+
202
+ ```
203
+
204
+
205
+
206
+ ## Example
207
+
208
+ <div style="display: flex; text-align: center; align-items: center; justify-content: space-between;">
209
+   <figure style="margin: 10px; width: calc(25% - 20px);">
210
+     <img src="examples/ref_image.png" alt="Ref_image" style="width: 100%;">
211
+     <figcaption>Ref_image</figcaption>
212
+   </figure>
213
+   <figure style="margin: 10px; width: calc(25% - 20px);">
214
+     <img src="examples/ref_mask.png" alt="Ref_mask" style="width: 100%;">
215
+     <figcaption>Ref_mask</figcaption>
216
+   </figure>
217
+   <figure style="margin: 10px; width: calc(25% - 20px);">
218
+     <img src="examples/tar_image.png" alt="Tar_image" style="width: 100%;">
219
+     <figcaption>Tar_image</figcaption>
220
+   </figure>
221
+   <figure style="margin: 10px; width: calc(25% - 20px);">
222
+     <img src="examples/tar_mask.png" alt="Tar_mask" style="width: 100%;">
223
+     <figcaption>Tar_mask</figcaption>
224
+   </figure>
225
+   <figure style="margin: 10px; width: calc(25% - 20px);">
226
+     <img src="examples/add.png" alt="Add" style="width: 100%;">
227
+     <figcaption>Add</figcaption>
228
+   </figure>
229
+   <figure style="margin: 10px; width: calc(25% - 20px);">
230
+     <img src="examples/replace.png" alt="Replace" style="width: 100%;">
231
+     <figcaption>Replace</figcaption>
232
+   </figure>
233
+ </div>
234
+
235
+ ### Text Prompt
236
+ Add Prompt: Add [label from `tar_image` (in label.json) ]</p>
237
+ Replace Prompt: Replace [label from `src_image` (in src_image/replace/replace_label.json) ] with [label from `tar_image` (in label.json) ]
238
+
239
+
240
+ ## Citation
241
+ ```
242
+ @article{song2025insert,
243
+ title={Insert Anything: Image Insertion via In-Context Editing in DiT},
244
+ author={Song, Wensong and Jiang, Hong and Yang, Zongxing and Quan, Ruijie and Yang, Yi},
245
+ journal={arXiv preprint arXiv:2504.15009},
246
+ year={2025}
247
+ }
248
+ ```