Preparing GLEAN Dataset

@InProceedings{chan2021glean,
  author = {Chan, Kelvin CK and Wang, Xintao and Xu, Xiangyu and Gu, Jinwei and Loy, Chen Change},
  title = {GLEAN: Generative Latent Bank for Large-Factor Image Super-Resolution},
  booktitle = {Proceedings of the IEEE conference on computer vision and pattern recognition},
  year = {2021}
}

Preparing cat_train dataset

1. Download cat dataset from LSUN homepage

2. Download cat_train/meta_info_LSUNcat_GT.txt from GLEAN homepage

3. Export and downsample images

Export images from lmdb file and resize the input images to the designated size. We provide such a script:

python tools/dataset_converters/glean/preprocess_cat_train_dataset.py --lmdb-path .data/cat --meta-file-path ./data/cat_train/meta_info_LSUNcat_GT.txt --out-dir ./data/cat_train

The generated data is stored under cat_train and the folder structure is as follows.

mmagic
├── mmagic
├── tools
├── configs
├── data
│   ├── cat_train
│   │   ├── GT
│   │   ├── BIx8_down
│   │   ├── BIx16_down
│   │   ├── meta_info_LSUNcat_GT.txt
...

Preparing cat_test dataset

1. Download CAT dataset from here.

2. Download cat_test/meta_info_Cat100_GT.txt from GLEAN homepage

3. Downsample images

Resize the input images to the designated size. We provide such a script:

python tools/dataset_converters/glean/preprocess_cat_test_dataset.py --data-path ./data/CAT_03 --meta-file-path ./data/cat_test/meta_info_Cat100_GT.txt --out-dir ./data/cat_test

The generated data is stored under cat_test and the folder structure is as follows.

mmagic
├── mmagic
├── tools
├── configs
├── data
│   ├── cat_test
│   │   ├── GT
│   │   ├── BIx8_down
│   │   ├── BIx16_down
│   │   ├── meta_info_Cat100_GT.txt
...

Preparing FFHQ dataset

1. Download FFHQ dataset (images1024x1024) from it’s homepage

Then you can refactor the folder structure looks like:

ffhq
├── images
|   ├── 00000.png
|   ├── 00001.png
|   ├── ...
|   ├── 69999.png

2. Download ffhq/meta_info_FFHQ_GT.txt from GLEAN homepage

3. Downsample images

Resize the input images to the designated size. We provide such a script:

python tools/dataset_converters/glean/preprocess_ffhq_celebahq_dataset.py --data-root ./data/ffhq/images

The generated data is stored under ffhq and the folder structure is as follows.

mmagic
├── mmagic
├── tools
├── configs
├── data
|   ├── ffhq
|   |   ├── images
│   │   ├── BIx8_down
|   |   ├── BIx16_down
|   |   ├── meta_info_FFHQ_GT.txt
...

Preparing CelebA-HQ dataset

1. Preparing datasets following it’s homepage

Then you can refactor the folder structure looks like:

CelebA-HQ
├── GT
|   ├── 00000.png
|   ├── 00001.png
|   ├── ...
|   ├── 30000.png

2. Download CelebA-HQ/meta_info_CelebAHQ_val100_GT.txt from GLEAN homepage

3. Downsample images

Resize the input images to the designated size. We provide such a script:

python tools/dataset_converters/glean/preprocess_ffhq_celebahq_dataset.py --data-root ./data/CelebA-HQ/GT

The generated data is stored under CelebA-HQ and the folder structure is as follows.

mmagic
├── mmagic
├── tools
├── configsdata
├── data
|   ├── CelebA-HQ
|   |   ├── GT
│   │   ├── BIx8_down
|   |   ├── BIx16_down
|   |   ├── meta_info_CelebAHQ_val100_GT.txt
...

Preparing FFHQ_CelebAHQ dataset

We merge FFHQ(ffhq/images) and CelebA-HQ(CelebA-HQ/GT) to generate FFHQ_CelebAHQ dataset.

The folder structure should looks like:

mmagic
├── mmagic
├── tools
├── configs
├── data
|   ├── FFHQ_CelebAHQ
|   |   ├── GT
...