The dataset consists of authentic images sampled from the Shutterstock platform across various categories, including a balanced selection where one-third of the images feature humans. These authentic images are paired with their equivalents generated using state-of-the-art generative models. This structured pairing enables a direct comparison between real and AI-generated content, providing a robust foundation for developing and evaluating image authenticity detection systems.

## The dataset with the images can be downloaded from [here](https://kaggle.com/datasets/97c52dbae208af8189b94d549ab0f2dfd90c3f5b14700d8506ba358d1a2d5c6f)

The dataset is composed of train_data and test_data folders with the images identified by the respective `id`. Alternatively, you have been provided a `train.csv` and `test.csv` files with the id and label of the train and test data.

The results you'll be submitting should correspond to the same order of the `test.csv` file `id`s and must not contain any other columns except for the `id` and `label` (predicted label)
## Files

*   **train.csv** - the training set (79,950 images)
*   **test.csv** - the test set (19,986 images)

## Columns

*   `id` - `train_data/a6dcb93f596a43249135678dfcfc17ea.jpg`
*   `label` - `1`