# Welcome to GenImageNet
This [OSF repository](https://osf.io/8ctjy/) hosts the AI-generated images from the paper:
[The power of generative marketing: Can generative AI create superhuman visual marketing content?](https://doi.org/10.1016/j.ijresmar.2024.09.002).
## Overview of the dataset
The dataset includes a total of 10,320 AI-generated images:
| | DALL-E 3 | Firefly 2 | Imagen 2 | Imagine | Midjourney v6 | Realistic Vision | SDXL Turbo | Total |
|-------------------|-----------|-----------|----------|---------|---------------|------------------|------------|------------|
| amazon | 300 | 30 | 30 | 30 | 300 | 300 | 300 | 1,290 |
| booking_com | 300 | 30 | 30 | 30 | 300 | 300 | 300 | 1,290 |
| image_ads | 300 | 30 | 30 | 30 | 300 | 300 | 300 | 1,290 |
| instagram | 300 | 30 | 30 | 30 | 300 | 300 | 300 | 1,290 |
| lions_awards | 300 | 30 | 30 | 30 | 300 | 300 | 300 | 1,290 |
| twitter | 300 | 30 | 30 | 30 | 300 | 300 | 300 | 1,290 |
| unsplash | 300 | 30 | 30 | 30 | 300 | 300 | 300 | 1,290 |
| yelp | 300 | 30 | 30 | 30 | 300 | 300 | 300 | 1,290 |
| **Total** | **2,400** | **240** | **240** | **240** | **2,400** | **2,400** | **2,400** | **10,320** |
The CSV and XLSX files **GenImageNet_imageLevel** and **GenImageNet_assignmentLevel** include the following variables per image and assignmentId, respectively:
| Variable | Description |
|---------------------|------------------------------------------------------------------------------|
| aestheticsScore | Neural Image Assessment (NIMA) score |
| assignmentId | Unique ID of a task completed by a specific workerId |
| colorSaturation | Sum of mean and standard deviation of S values in HSV color space |
| dataset | Origin of the human-made source image |
| filename | Filename of the AI-generated image (see below for naming convention) |
| hasFace | Binary flag to indicate if a face is displayed on the image (via MTCNN) |
| hasText | Binary flag to indicate if text is displayed on the image (via tesseract) |
| hasVisibleWatermark | Binary flag to indicate if a visual watermark is visible on the image |
| model | AI model used for image generation |
| prompt | Input prompt used to generate the image with the given AI model |
| qualityScore | 7-point Likert rating for perceived quality for the given assignmentId |
| realismScore | 7-point Likert rating for perceived realism for the given assignmentId |
| visualComplexity | File size in kilobyte divided by the image resolution (i.e., width × height) |
| workerId | Unique ID per worker/rater |
## Naming convention for filenames
Filenames are coded as follows: <_model_id_>\_<_dataset_id_>\_<_image_incrementor_>.webp
The assoicated keys are:
| _model_id_ | model_name |
|----------|------------------|
| 1 | DALL-E 3 |
| 2 | Midjourney v6 |
| 3 | Firefly 2 |
| 4 | Imagen 2 |
| 5 | Imagine |
| 6 | Realistic Vision |
| 7 | SDXL Turbo |
| _dataset_id_ | dataset_name |
|------------|----------------|
| a | Amazon |
| b | Booking.com |
| c | Image ads |
| d | Instagram |
| e | Lions awards |
| f | Twitter |
| g | Unsplash |
| h | Yelp |
_image_incrementor_ is a 4 digit integer value
### Examples:
- _1_b_1048.webp_ is an image generated by _DALL-E 3_ based on an image from the _Booking.com_ dataset.
- _6_a_0029.webp_ is an image generated by _Realistic Vision_ based on an image from the _Amazon_ dataset.
# License
This dataset is published under **CC BY-NC-ND 4.0**
For more information see: https://creativecommons.org/licenses/by-nc-nd/4.0/
Please use the following citation when using this dataset:
J. Hartmann, Y. Exner, S. Domdey, The power of generative marketing: Can generative AI create superhuman visual marketing content?, International Journal of Research in Marketing (2024), doi: https://doi.org/10.1016/j.ijresmar.2024.09.002
