Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -3,4 +3,44 @@ license: mit
|
|
3 |
language:
|
4 |
- en
|
5 |
pretty_name: W-Bench
|
6 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
language:
|
4 |
- en
|
5 |
pretty_name: W-Bench
|
6 |
+
size: 10,000 instances
|
7 |
+
---
|
8 |
+
|
9 |
+
# What is it?
|
10 |
+
W-Bench is the first holistic benchmark that incorporates four types of image editing techniques to assess the robustness of watermarking methods. Eleven representative watermarking methods are evaluated on the W-Bench. The W-Bench contains 10,000 images sourced from datasets such as COCO, Flickr, ShareGPT4V, etc.
|
11 |
+
|
12 |
+
# Dataset Structure
|
13 |
+
|
14 |
+
The evaluation set is divided into 6 different categories:
|
15 |
+
- 1,000 samples for stochastic regeneration
|
16 |
+
- 1,000 samples for deterministic regeneration
|
17 |
+
- 1,000 samples for global editing
|
18 |
+
- 5,000 samples for local editing (divided into five sets, each containing 1,000 images, with different mask sizes ranging from 10–60% of the image area)
|
19 |
+
- 1,000 samples for image-to-video generation
|
20 |
+
- 1,000 samples for testing conventional distortion
|
21 |
+
|
22 |
+
# How to download and use 🍷 W-Bench
|
23 |
+
|
24 |
+
## Using `huggingface_hub`
|
25 |
+
|
26 |
+
```python
|
27 |
+
from huggingface_hub import snapshot_download
|
28 |
+
folder = snapshot_download(
|
29 |
+
"Shilin-LU/W-Bench",
|
30 |
+
repo_type="dataset",
|
31 |
+
local_dir="./W-Bench/",
|
32 |
+
allow_patterns="DET_INVERSION_1K/image/*")
|
33 |
+
```
|
34 |
+
|
35 |
+
For faster downloads, make sure to install `pip install huggingface_hub[hf_transfer]` and set the environment variable `HF_HUB_ENABLE_HF_TRANSFER=1`.
|
36 |
+
|
37 |
+
## Using `datasets`
|
38 |
+
|
39 |
+
```python
|
40 |
+
from datasets import load_dataset
|
41 |
+
wbench = load_dataset("Shilin-LU/W-Bench", streaming=True)
|
42 |
+
```
|
43 |
+
|
44 |
+
# Citation Information
|
45 |
+
Paper on [arXiv](https://arxiv.org/abs/2410.18775)
|
46 |
+
|