OpenSDI: Spotting Diffusion-Generated Images in the Open World
1
45 commits
1 linked in READMEs
updated Mar 26, 2025
README
OpenSDI: Spotting Diffusion-Generated Images in the Open World
This dataset is designed to address the OpenSDI challenge: spotting diffusion-generated images in realistic, open-world scenarios. It is described in the paper:
Project Page: https://iamwangyabin.github.io/OpenSDI/
OpenSDID Dataset Highlights:
- User Diversity: Simulates a wide range of user intentions and creative styles using diverse text prompts generated by VLMs.
- Model Innovation: Includes images from multiple state-of-the-art diffusion models (SD1.5, SD2.1, SDXL, SD3, Flux.1) released after 2023.
- Manipulation Scope: Covers the full spectrum of diffusion-based manipulation, from global image synthesis to precise local edits, object insertions, and background changes.
Dataset Overview:
| Model | Training Set | Test Set | Total | ||
|---|---|---|---|---|---|
| Real | Fake | Real | Fake | Images | |
| SD1.5 | 100K | 100K | 10K | 10K | 220K |
| SD2.1 | - | - | 10K | 10K | 20K |
| SDXL | - | - | 10K | 10K | 20K |
| SD3 | - | - | 10K | 10K | 20K |
| Flux.1 | - | - | 10K | 10K | 20K |
| Total | 100K | 100K | 50K | 50K | 300K |
Download:
- OpenSDI Training Dataset: https://huggingface.co/datasets/nebula/OpenSDI_train
- OpenSDI Testing Dataset: This repository
This dataset is for academic use and is distributed under the CC BY-SA 4.0 license. Real images come from megalith-10m.
OpenSDI: Spotting Diffusion-Generated Images in the Open World
1
45 commits
1 linked in READMEs
updated Mar 26, 2025
README
OpenSDI: Spotting Diffusion-Generated Images in the Open World
This dataset is designed to address the OpenSDI challenge: spotting diffusion-generated images in realistic, open-world scenarios. It is described in the paper:
Project Page: https://iamwangyabin.github.io/OpenSDI/
OpenSDID Dataset Highlights:
- User Diversity: Simulates a wide range of user intentions and creative styles using diverse text prompts generated by VLMs.
- Model Innovation: Includes images from multiple state-of-the-art diffusion models (SD1.5, SD2.1, SDXL, SD3, Flux.1) released after 2023.
- Manipulation Scope: Covers the full spectrum of diffusion-based manipulation, from global image synthesis to precise local edits, object insertions, and background changes.
Dataset Overview:
| Model | Training Set | Test Set | Total | ||
|---|---|---|---|---|---|
| Real | Fake | Real | Fake | Images | |
| SD1.5 | 100K | 100K | 10K | 10K | 220K |
| SD2.1 | - | - | 10K | 10K | 20K |
| SDXL | - | - | 10K | 10K | 20K |
| SD3 | - | - | 10K | 10K | 20K |
| Flux.1 | - | - | 10K | 10K | 20K |
| Total | 100K | 100K | 50K | 50K | 300K |
Download:
- OpenSDI Training Dataset: https://huggingface.co/datasets/nebula/OpenSDI_train
- OpenSDI Testing Dataset: This repository
This dataset is for academic use and is distributed under the CC BY-SA 4.0 license. Real images come from megalith-10m.