Duplicate from HPAI-BSC/SuSy-Dataset

Co-authored-by: Pablo Bernabeu <pabberpe@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,15 @@
+# Track zip files with Git LFS
+*.zip filter=lfs diff=lfs merge=lfs -text
+
+# Track zip parts with Git LFS
+*.z01 filter=lfs diff=lfs merge=lfs -text
+*.z02 filter=lfs diff=lfs merge=lfs -text
+*.z03 filter=lfs diff=lfs merge=lfs -text
+*.z04 filter=lfs diff=lfs merge=lfs -text
+*.z05 filter=lfs diff=lfs merge=lfs -text
+*.z06 filter=lfs diff=lfs merge=lfs -text
+*.z07 filter=lfs diff=lfs merge=lfs -text
+*.z08 filter=lfs diff=lfs merge=lfs -text
+*.z09 filter=lfs diff=lfs merge=lfs -text
+*.z10 filter=lfs diff=lfs merge=lfs -text
+*.z11 filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,228 @@
+---
+pretty_name: SuSy Dataset
+task_categories:
+- image-classification
+size_categories:
+- 10K<n<100K
+tags:
+- image
+- ai-images
+- synthetic-image-detection
+configs:
+- config_name: susy_dataset
+  data_files:
+  - split: train
+    path: data/train.zip
+  - split: val
+    path: data/val.zip
+  - split: test
+    path: data/test.zip
+dataset_info:
+  features:
+    - name: image
+      dtype: image
+    - name: label
+      dtype:
+        class_label:
+          names:
+            '0': coco
+            '1': dalle-3-images
+            '2': diffusiondb
+            '3': midjourney-images
+            '4': midjourney_tti
+            '5': realisticSDXL
+---
+
+<div align="center">
+  <h2>SuSy Dataset: Synthetic Image Detection</h2>  
+    The SuSy Dataset is a collection of authentic and synthetic images intended for training and evaluating synthetic content detectors. It was originally curated to train <a href="https://huggingface.co/HPAI-BSC/SuSy">SuSy</a> but can be used for any synthetic image detector model. This dataset is presented and used in the paper "<a href="https://arxiv.org/abs/2409.14128">Present and Future Generalization of Synthetic Image Detectors</a>".
+  <img src="https://cdn-uploads.huggingface.co/production/uploads/620683e7eeb1b73d904c96e5/Oy8RlHv9WuiznSpBct_d4.png" alt="image" width="300" height="auto">
+</div>
+<hr>
+<div align="center" style="line-height: 1;">
+  <a href="https://arxiv.org/abs/2409.14128" target="_blank" style="margin: 2px;">
+    <img alt="Paper" src="https://img.shields.io/badge/arXiv-2409.14128-b31b1b.svg" style="display: inline-block; vertical-align: middle;"/>
+  </a>
+  <a href="https://github.com/HPAI-BSC/SuSy" target="_blank" style="margin: 2px;">
+    <img alt="Repository" src="https://img.shields.io/badge/Repository-GitHub-181717?logo=github&logoColor=white" style="display: inline-block; vertical-align: middle;"/>
+  </a>
+  <a href="https://huggingface.co/HPAI-BSC/SuSy" target="_blank" style="margin: 2px;">
+    <img alt="Model" src="https://img.shields.io/badge/Model-Hugging%20Face-FFD21E?logo=huggingface" style="display: inline-block; vertical-align: middle;"/>
+  </a>
+</div>
+
+<div align="center" style="line-height: 1;">
+  <a href="https://huggingface.co/spaces/HPAI-BSC/SuSyGame" target="_blank" style="margin: 2px;">
+    <img alt="SuSy Challenge" src="https://img.shields.io/badge/SuSy%20Challenge-Hugging%20Face-FFD21E?logo=huggingface" style="display: inline-block; vertical-align: middle;"/>
+  </a>
+  <a href="https://huggingface.co/spaces/HPAI-BSC/SuSy" target="_blank" style="margin: 2px;">
+    <img alt="Interactive Demo" src="https://img.shields.io/badge/Interactive%20Demo-Hugging%20Face-FFD21E?logo=huggingface" style="display: inline-block; vertical-align: middle;"/>
+  </a>
+  <a href="https://colab.research.google.com/drive/15nxo0FVd-snOnj9TcX737fFH0j3SmS05" target="_blank" style="margin: 2px;">
+    <img alt="Code Demo" src="https://img.shields.io/badge/Code%20Demo-Colab-F9AB00?logo=googlecolab&logoColor=white" style="display: inline-block; vertical-align: middle;"/>
+  </a>
+</div>
+
+<div align="center" style="line-height: 1;">
+  <a href="https://hpai.bsc.es/" target="_blank" style="margin: 2px;">
+    <img alt="HPAI Website" src="https://img.shields.io/badge/HPAI-Website-blue" style="display: inline-block; vertical-align: middle;"/>
+  </a>
+  <a href="https://www.linkedin.com/company/hpai" target="_blank" style="margin: 2px;">
+    <img alt="LinkedIn" src="https://custom-icon-badges.demolab.com/badge/LinkedIn-0A66C2?logo=linkedin-white&logoColor=fff" style="display: inline-block; vertical-align: middle;"/>
+  </a>
+  <a href="https://bsky.app/profile/hpai.bsky.social" target="_blank" style="margin: 2px;">
+    <img alt="Bluesky" src="https://img.shields.io/badge/Bluesky-0285FF?logo=bluesky&logoColor=fff" style="display: inline-block; vertical-align: middle;"/>
+  </a>
+</div>
+
+**Image Examples**
+
+| COCO | dalle-3-images | diffusiondb |
+|:----:|:--------------:|:-----------:|
+| <img src="https://cdn-uploads.huggingface.co/production/uploads/620683e7eeb1b73d904c96e5/pbblFS9FmtQjBpcmpKSr3.jpeg" alt="image" width="300" height="auto"> | <img src="https://cdn-uploads.huggingface.co/production/uploads/620683e7eeb1b73d904c96e5/M1qin2gFq0ncYhqn3e7bK.jpeg" alt="image" width="300" height="auto"> | <img src="https://cdn-uploads.huggingface.co/production/uploads/620683e7eeb1b73d904c96e5/Zee5lKGpC62MKFKzZ49qb.png" alt="image" width="300" height="auto"> |
+
+| midjourney-images | midjourney-tti | realisticSDXL |
+|:-----------------:|:--------------:|:-------------:|
+| <img src="https://cdn-uploads.huggingface.co/production/uploads/620683e7eeb1b73d904c96e5/cG5_as0Dfa7TsE3RzCDyc.jpeg" alt="image" width="300" height="auto"> | <img src="https://cdn-uploads.huggingface.co/production/uploads/620683e7eeb1b73d904c96e5/8JEGhXGnb3lvDs0kfqU4h.png" alt="image" width="300" height="auto"> | <img src="https://cdn-uploads.huggingface.co/production/uploads/620683e7eeb1b73d904c96e5/aE79Ldjc5dVUk7p_gp5eI.png" alt="image" width="300" height="auto"> |
+
+## Dataset Details
+
+### Dataset Description
+
+The SuSy Dataset is a curated collection of real and AI-generated images, collected for the training and evaluation of synthetic image detectors. It includes images from various sources to ensure diversity and representativeness.
+
+- **Curated by:** [Pablo Bernabeu Perez](https://huggingface.co/pabberpe)
+- **License:** Multiple licenses (see individual dataset details)
+
+## Uses
+
+### Direct Use
+
+This dataset is intended for:
+- Replicating experiments related to SuSy
+- Training synthetic image detection and attribution models
+- Evaluating synthetic image detection and attribution models
+
+### Out-of-Scope Use
+
+The SuSy Dataset is specifically designed for synthetic image detection, classification, and attribution tasks. Therefore, the following uses considered out-of-scope:
+
+- Generating synthetic images: This dataset should not be used as training data for generative models or any attempts to create synthetic images.
+- Image manipulation: The dataset should not be used to develop or train models for altering, enhancing, or manipulating images.
+- Legal or forensic analysis: The dataset is not designed for use in legal proceedings or forensic investigations related to image authenticity.
+- Content moderation: While the dataset contains both authentic and synthetic images, it is not intended for general content moderation purposes beyond synthetic image detection.
+
+## Dataset Structure
+
+The dataset consists of two main types of images:
+
+- Real-world images: Photographs from the COCO dataset
+- Synthetic images: AI-generated images from five different generators
+
+### Training Data
+
+|      Dataset      | Year | Train | Validation |  Test | Total |
+|:-----------------:|:----:|:-----:|:----------:|:-----:|:-----:|
+|        COCO       | 2017 | 2,967 |    1,234   | 1,234 | 5,435 |
+|   dalle-3-images  | 2023 |  987  |     330    |  330  | 1,647 |
+|    diffusiondb    | 2022 | 2,967 |    1,234   | 1,234 | 5,435 |
+|   realisticSDXL   | 2023 | 2,967 |    1,234   | 1,234 | 5,435 |
+|   midjourney-tti  | 2022 | 2,718 |     906    |  906  | 4,530 |
+| midjourney-images | 2023 | 1,845 |     617    |  617  | 3,079 |
+
+#### Authentic Images
+- [COCO](https://cocodataset.org/) (Common Objects in Context): A large-scale object detection, segmentation, and captioning dataset. It includes over 330,000 images, with 200,000 labeled using 80 object categories. For this dataset, we use a random subset of 5,435 images.
+  - **License:** Creative Commons Attribution 4.0 license
+
+#### Synthetic Images
+- [dalle-3-images](https://huggingface.co/datasets/ehristoforu/dalle-3-images): Contains 3,310 unique images generated using DALL-E 3. The dataset does not include the prompts used to generate the images.
+  - **License:** MIT license
+
+- [diffusiondb](https://poloclub.github.io/diffusiondb/): A large-scale text-to-image prompt dataset containing 14 million images generated by Stable Diffusion 1.x series models (2022). We use a random subset of 5,435 images.
+  - **License:** CC0 1.0 Universal license
+
+- [realisticSDXL](https://huggingface.co/datasets/DucHaiten/DucHaiten-realistic-SDXL): Contains images generated using the Stable Diffusion XL (SDXL) model released in July 2023. We use only the "realistic" category, which contains 5,435 images.
+  - **License:** CreativeML OpenRAIL-M license
+
+- [midjourney-tti](https://www.kaggle.com/datasets/succinctlyai/midjourney-texttoimage): Contains images generated using Midjourney V1 or V2 models (early 2022). The original dataset provided URLs, which were scraped to obtain the images.
+  - **License:** CC0 1.0 Universal license (for links only, images are property of users who generated them)
+
+- [midjourney-images](https://huggingface.co/datasets/ehristoforu/midjourney-images): Contains 4,308 unique images generated using Midjourney V5 and V6 models (2023).
+  - **License:** MIT license
+
+## Dataset Creation
+
+### Curation Rationale
+
+This dataset was created to provide a comprehensive set of both real and AI-generated images for training and evaluating synthetic content detectors. The curation process aimed to:
+
+- Include diverse and high-quality data from multiple sources
+- Represent various AI image generation models (DALL-E, Midjourney, Stable Diffusion)
+- Include both early (2022) and more recent (2023) AI-generated images to study the impact of model evolution
+
+### Source Data
+
+#### Data Collection and Processing
+
+- COCO and diffusiondb datasets were undersampled to 5,435 images each to balance with other datasets.
+- The realisticSDXL dataset uses only the "realistic" category images.
+- Existing train, validation, and test partitions are respected where available.
+- For datasets without predefined splits, a 60%-20%-20% random split is performed for train, validation and test sets respectively.
+- The midjourney-tti dataset had collage images and mosaics removed.
+- The dalle-3-images and midjourney-images datasets were deduplicated.
+
+#### Who are the source data producers?
+
+- Real-world images: Photographers (COCO dataset)
+- Synthetic images: Various AI image generation models (DALL-E, Stable Diffusion and Midjourney)
+
+## Bias, Risks, and Limitations
+
+- The dataset may not fully represent the entire spectrum of real-world or AI-generated images.
+- The performance of models trained on this dataset may vary depending on the specific characteristics of each subset.
+- As AI image generation technology rapidly evolves, the synthetic images in this dataset may become less representative of current AI capabilities over time.
+
+### Recommendations
+
+Users should be aware that:
+- This dataset contains both real and AI-generated images, each subject to its own license depending on the dataset of origin.
+- The dataset is specifically curated for synthetic image detection and attribution, and may not be suitable for other computer vision tasks without modification.
+- When using this dataset, proper attribution should be given to the original sources as per their respective licenses.
+- Regular updates to the dataset may be necessary to keep pace with advancements in AI image generation technology.
+
+## More Information
+
+For more detailed information about the dataset composition and the SuSy model, please refer to the original [research paper](https://arxiv.org/abs/2409.14128).
+
+**BibTeX:**
+
+```bibtex
+@misc{bernabeu2024susy,
+    title={Present and Future Generalization of Synthetic Image Detectors}, 
+    author={Pablo Bernabeu-Perez and Enrique Lopez-Cuena and Dario Garcia-Gasulla},
+    year={2024},
+    eprint={2409.14128},
+    archivePrefix={arXiv},
+    primaryClass={cs.CV},
+    url={https://arxiv.org/abs/2409.14128}, 
+}
+
+```
+
+```bibtex
+@thesis{bernabeu2024aidetection,
+    title={Detecting and Attributing AI-Generated Images with Machine Learning},
+    author={Bernabeu Perez, Pablo},
+    school={UPC, Facultat d'Informàtica de Barcelona, Departament de Ciències de la Computació},
+    year={2024},
+    month={06}
+}
+```
+
+## Model Card Authors
+
+[Pablo Bernabeu Perez](https://huggingface.co/pabberpe)
+
+## Model Card Contact
+
+For further inquiries, please contact [HPAI](mailto:hpai@bsc.es)

data/test.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cf4d744686274ce7b7624f83cb04247792d6ad48d1f7d4e944faa880ebfe242
+size 6020979522

data/train.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c349404c2f2fff72348ace1c450b92e5bdf1a9d6508a5c6cc38bbfe002ecdc8
+size 15189567725

data/val.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c41beef0811e5dd70db3923290ee1fc57a73eb312bef9de5ad573850578775ff
+size 4725427667