Dataset Description
=============
* Dataset Copyright
  * All data is generated from the dataset provided by [Faceforensics](https://github.com/ondyari/FaceForensics).
  * We don't have any ownership of this data, nor do we have participants in this contest.
* Format of Dataset
 * Inside the dataset, **the images have different sizes**. You should manage the input image by yourself.
 * Every image in this dataset has channel order as follows
    * RGB

![RGB format image](https://raw.githubusercontent.com/nkmk/python-snippets/7b135319354f28ed55244fae00a7695a2c14e372/notebook/data/dst/lena_bgr_cv.jpg)    
    * BGR

![BGR format image](https://raw.githubusercontent.com/nkmk/python-snippets/7b135319354f28ed55244fae00a7695a2c14e372/notebook/data/dst/lena_bgr_pillow.jpg)

* Dataset ratio
    * Training dataset
         * Total 10000 (Real:5000 counts ,Fake:5000 counts )
         * Real: Real image from Faceforensics dataset
         * Fake: Synthetic image that generated by Neural Texture and input was Real Faceforensics dataset    
    * Validation dataset
        * Total 5000 ( Real: 2500 counts, Fake:2500 counts)
        * Real: Real image from Faceforensics dataset
        * Fake: Synthetic image that generated by Neural Texture and input was Real Faceforensics     
    * Test dataset
        * Total 2000 (Unknown Real, Fake ratio)
        * Real: Real image from Faceforensics dataset
        * Fake: Synthetic image that generated by Neural Texture and input was Real Faceforensics