Commit
·
4d5f0c4
1
Parent(s):
0649a8c
Update README.md
Browse files
README.md
CHANGED
|
@@ -17,8 +17,50 @@ dataset_info:
|
|
| 17 |
num_bytes: 15474910323.726
|
| 18 |
num_examples: 8807
|
| 19 |
- name: dev
|
| 20 |
-
num_bytes: 926910159
|
| 21 |
num_examples: 528
|
| 22 |
download_size: 16474461666
|
| 23 |
dataset_size: 16401820482.726
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
num_bytes: 15474910323.726
|
| 18 |
num_examples: 8807
|
| 19 |
- name: dev
|
| 20 |
+
num_bytes: 926910159
|
| 21 |
num_examples: 528
|
| 22 |
download_size: 16474461666
|
| 23 |
dataset_size: 16401820482.726
|
| 24 |
+
task_categories:
|
| 25 |
+
- text-to-image
|
| 26 |
+
language:
|
| 27 |
+
- en
|
| 28 |
+
pretty_name: MagicBrush
|
| 29 |
+
size_categories:
|
| 30 |
+
- 10K<n<100K
|
| 31 |
---
|
| 32 |
+
|
| 33 |
+
|
| 34 |
+
# Dataset Card for Dataset Name
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
### Dataset Summary
|
| 38 |
+
|
| 39 |
+
MagicBrush is the first large-scale, manually-annotated instruction-guided image editing dataset covering diverse scenarios single-turn, multi-turn, mask-provided, and mask-free editing. MagicBrush comprises 10K (source image, instruction, target image) triples, which is sufficient to train large-scale image editing models.
|
| 40 |
+
|
| 41 |
+
Please check our [website](https://osu-nlp-group.github.io/MagicBrush/) to explore more visual results.
|
| 42 |
+
|
| 43 |
+
Dataset Structure
|
| 44 |
+
|
| 45 |
+
Data Fields
|
| 46 |
+
|
| 47 |
+
"img_id" (str): same from COCO id
|
| 48 |
+
"turn_index" (int32): the edit turn in the image
|
| 49 |
+
"input_img" (str): input image, could be the original real image (turn_index=1) and edited images from last tunr (turn_index >=2)
|
| 50 |
+
"instruction" (str): edit instruction of how the input image should be changed.
|
| 51 |
+
"output_img" (str): the edited image corresponding to the input image and instruction.
|
| 52 |
+
|
| 53 |
+
Data Splits
|
| 54 |
+
|
| 55 |
+
train: 8,807 edit turns (4,512 edit sessions).
|
| 56 |
+
val: 528 edit turns (266 edit sessions).
|
| 57 |
+
test: (To prevent potential data leakage, please check our repo for information on obtaining the test set.)
|
| 58 |
+
Licensing Information
|
| 59 |
+
|
| 60 |
+
Creative Commons License
|
| 61 |
+
This work is licensed under a Creative Commons Attribution 4.0 International License.
|
| 62 |
+
|
| 63 |
+
|
| 64 |
+
Citation Information
|
| 65 |
+
|
| 66 |
+
TODO
|