Update README.md
Browse files
README.md
CHANGED
@@ -1,5 +1,41 @@
|
|
1 |
-
---
|
2 |
-
license: other
|
3 |
-
license_name: pi-lab-license-1.0
|
4 |
-
license_link: LICENSE
|
5 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: other
|
3 |
+
license_name: pi-lab-license-1.0
|
4 |
+
license_link: LICENSE
|
5 |
+
---
|
6 |
+
|
7 |
+
# This is the official datasets for the paper: FaceMe: Robust Blind Face Restoration With Personal Identification
|
8 |
+
|
9 |
+
|
10 |
+
## Dataset Overview
|
11 |
+
|
12 |
+
This repository contains the following five datasets:
|
13 |
+
|
14 |
+
1. **FFHQ**
|
15 |
+
2. **FFHQRef**
|
16 |
+
3. **LFW-Test-Ref**
|
17 |
+
4. **Wider-Test-Ref**
|
18 |
+
5. **WebPhoto-Test-Ref**
|
19 |
+
|
20 |
+
### Dataset Usage
|
21 |
+
|
22 |
+
- **FFHQ** and **FFHQRef**: Used for training.
|
23 |
+
- **LFW-Test-Ref**, **Wider-Test-Ref**, and **WebPhoto-Test-Ref**: Used for inference.
|
24 |
+
|
25 |
+
### Important Notes
|
26 |
+
|
27 |
+
Due to the large size of the **FFHQRef** dataset, it has been split into multiple parts. Before using this dataset, you need to merge the split files into a single `.tar` file and then extract it.
|
28 |
+
|
29 |
+
#### Steps to Merge the Split Files
|
30 |
+
|
31 |
+
Run the following command to merge all parts into a single file:
|
32 |
+
|
33 |
+
```bash
|
34 |
+
ls FFHQRef_part_*.tar | sort -V | xargs cat > merged_FFHQRef.tar
|
35 |
+
```
|
36 |
+
|
37 |
+
After merging, extract the merged_FFHQRef.tar file to access the dataset:
|
38 |
+
|
39 |
+
```bash
|
40 |
+
tar -xvf merged_FFHQRef.tar
|
41 |
+
```
|