| # FLUX Kontext Object Placement Dataset | |
| A paired dataset for training object placement and image editing models using FLUX.1-Kontext architecture. | |
| ## Dataset Overview | |
| This dataset contains 503 images organized into paired source and target images for object placement tasks. | |
| Constructed using https://github.com/google/navi/ | |
| ### Dataset Structure | |
| ``` | |
| images/ | |
| ├── start/ # Source images (250 images, 64MB) | |
| ├── end/ # Target images (251 images, 498MB) | |
| └── validation/ # Validation images (3 images, 588KB) | |
| ``` | |
| ### Dataset Statistics | |
| - **Total Images**: 503 | |
| - **Training Pairs**: 250 paired images | |
| - **Validation Images**: 3 | |
| ## Related Resources | |
| - **Trained Model**: [fediry/flux-kontext-object-placement-lora](https://huggingface.co/fediry/flux-kontext-object-placement-lora) | |
| - **Base Model**: [black-forest-labs/FLUX.1-Kontext-dev](https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev) | |
| - **Training Framework**: [AI-Toolkit](https://github.com/ostris/ai-toolkit) | |