Commit
·
d50478c
1
Parent(s):
78a7c9c
Added description.
Browse files
README.md
CHANGED
|
@@ -1,3 +1,91 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# MIT-Adobe FiveK Dataset
|
| 2 |
+
|
| 3 |
+
The MIT-Adobe FiveK Dataset [[1]]( #references ) is a publicly available dataset providing the following items.
|
| 4 |
+
1. 5,000 RAW images in DNG format
|
| 5 |
+
2. retouched images of each RAW image by five experts in TIFF format (25,000 images, 16 bits per channel, ProPhoto RGB color space, and lossless compression)
|
| 6 |
+
3. semantic information about each image
|
| 7 |
+
The dataset was created by MIT and Adobe Systems, Inc., and is intended to provide a diverse and challenging set of images for testing image processing algorithms. The images were selected to represent a wide range of scenes, including landscapes, portraits, still lifes, and architecture. The images also vary in terms of lighting conditions, color balance, and exposure.
|
| 8 |
+
In practice, this dataset is often used after RAW images have undergone various processing steps. For example, RAW images are developed by adding noise, overexposure, and underexposure to emulate camera errors.
|
| 9 |
+
However, the officially provided dataset has a complex structure and is difficult to handle. This repository provides tools to easily download and use the datasets.
|
| 10 |
+
|
| 11 |
+
## Official Website
|
| 12 |
+
|
| 13 |
+
[MIT-Adobe FiveK Dataset](https://data.csail.mit.edu/graphics/fivek/)
|
| 14 |
+
|
| 15 |
+
## License
|
| 16 |
+
|
| 17 |
+
- [LicenseAdobe.txt](https://data.csail.mit.edu/graphics/fivek/legal/LicenseAdobe.txt) covers files listed in [filesAdobe.txt](https://data.csail.mit.edu/graphics/fivek/legal/filesAdobe.txt)
|
| 18 |
+
- [LicenseAdobeMIT.txt](https://data.csail.mit.edu/graphics/fivek/legal/LicenseAdobeMIT.txt) covers files listed in [filesAdobeMIT.txt](https://data.csail.mit.edu/graphics/fivek/legal/filesAdobeMIT.txt)
|
| 19 |
+
|
| 20 |
+
## Data Samples
|
| 21 |
+
|
| 22 |
+
|Raw (DNG)|Expert A|Expert B|Expert C|Expert D|Expert E|Categories|Camera Model|
|
| 23 |
+
|---|---|---|---|---|---|---|---|
|
| 24 |
+
|[a0001-jmac_</br >DSC1459.dng](https://data.csail.mit.edu/graphics/fivek/img/dng/a0001-jmac_DSC1459.dng)||||||{"location":"outdoor","time": "day","light": "sun_sky","subject": "nature"}|Nikon D70|
|
| 25 |
+
|[a1384-dvf_095.dng](https://data.csail.mit.edu/graphics/fivek/img/dng/a1384-dvf_095.dng)||||||{ "location": "outdoor", "time": "day", "light": "sun_sky", "subject": "nature" }|Leica M8|
|
| 26 |
+
|[a4607-050801_</br >080948__</br >I2E5512.dng](https://data.csail.mit.edu/graphics/fivek/img/dng/a4607-050801_080948__I2E5512.dng)||||||{ "location": "indoor", "time": "day", "light": "artificial", "subject": "people" }|Canon EOS-1D Mark II|
|
| 27 |
+
|
| 28 |
+
# References
|
| 29 |
+
|
| 30 |
+
```
|
| 31 |
+
@inproceedings{fivek,
|
| 32 |
+
author = "Vladimir Bychkovsky and Sylvain Paris and Eric Chan and Fr{\'e}do Durand",
|
| 33 |
+
title = "Learning Photographic Global Tonal Adjustment with a Database of Input / Output Image Pairs",
|
| 34 |
+
booktitle = "The Twenty-Fourth IEEE Conference on Computer Vision and Pattern Recognition",
|
| 35 |
+
year = "2011"
|
| 36 |
+
}
|
| 37 |
+
```
|
| 38 |
+
|
| 39 |
+
# Code
|
| 40 |
+
|
| 41 |
+
[GitHub repository](https://github.com/yuukicammy/mit-adobe-fivek-dataset) provides tools to download and use MIT-Adobe FiveK Dataset in a machine learning friendly manner.
|
| 42 |
+
You can download the dataset with a single line of Python code. Also, you can use Pytorch's DetaLoader to iteratively retrieve data for your own use.
|
| 43 |
+
The processing can be easily accomplished with multiprocessing with Pytorch's DataLoader!
|
| 44 |
+
|
| 45 |
+
## Requirements
|
| 46 |
+
- Python 3.7 or greater
|
| 47 |
+
- Pytorch 2.X
|
| 48 |
+
- tqdm
|
| 49 |
+
- urllib3
|
| 50 |
+
|
| 51 |
+
## Usage
|
| 52 |
+
|
| 53 |
+
You can use as follows.
|
| 54 |
+
|
| 55 |
+
<span style="color:red">
|
| 56 |
+
NOTE: For DataLoader, MUST set `batch_size` to `None` to disable automatic batching.
|
| 57 |
+
</span>
|
| 58 |
+
|
| 59 |
+
```python
|
| 60 |
+
from torch.utils.data.dataloader import DataLoader
|
| 61 |
+
from dataset.fivek import MITAboveFiveK
|
| 62 |
+
|
| 63 |
+
metadata_loader = DataLoader(
|
| 64 |
+
MITAboveFiveK(root="path-to-dataset-root", split="train", download=True, experts=["a"]),
|
| 65 |
+
batch_size=None, num_workers=2)
|
| 66 |
+
|
| 67 |
+
for item in metadata_loader:
|
| 68 |
+
# Processing as you want.
|
| 69 |
+
# Add noise, overexpose, underexpose, etc.
|
| 70 |
+
print(item["files"]["dng"])
|
| 71 |
+
```
|
| 72 |
+
|
| 73 |
+
## Example
|
| 74 |
+
|
| 75 |
+
Please see [sample code](https://github.com/yuukicammy/mit-adobe-fivek-dataset/blob/master/sample_process.py) .
|
| 76 |
+
|
| 77 |
+
## API
|
| 78 |
+
|
| 79 |
+
CLASS MITAboveFiveK(torch.utils.data.dataset.Dataset)
|
| 80 |
+
- - -
|
| 81 |
+
MITAboveFiveK(root: str, split: str, download: bool = False, experts: List[str] = None) -> None
|
| 82 |
+
|
| 83 |
+
- root (str):
|
| 84 |
+
The root directory where the MITAboveFiveK directory exists or to be created.
|
| 85 |
+
- split (str):
|
| 86 |
+
One of {'train', 'val', 'test', 'debug'}. 'debug' uses only 9 data contained in 'train'.
|
| 87 |
+
- download (bool):
|
| 88 |
+
If True, downloads the dataset from the official urls. Files that already exist locally will skip the download. Defaults to False.
|
| 89 |
+
- experts (List[str]):
|
| 90 |
+
List of {'a', 'b', 'c', 'd', 'e'}. 'a' means 'Expert A' in the [website](https://data.csail.mit.edu/graphics/fivek/ ). If None or empty list, no expert data is used. Defaults to None.
|
| 91 |
+
|