CreatiLayout: Layout-to-Image generation
CreatiLayout is a layout-to-image framework for Diffusion Transformer models, offering high-quality and fine-grained controllable generation based on the global description and entity annotations. Users need to provide a global description and the position and description of each entity, as shown in the examples. Please feel free to modify the position and attributes of the entities in the examples (such as size, color, shape, text, portrait, etc.). Here are some inspirations: Iron Man -> Spider Man/Harry Potter/Buzz Lightyear; CreatiLayout -> Hello Friends/Let's Control; drawing board -> round drawing board; Modify the position of the drawing board to (0.4, 0.15, 0.55, 0.35)
Region Captions
Region Captions |
|---|
Region Bounding Boxes (x_min,y_min,x_max,y_max)
x1 | y1 | x2 | y2 |
|---|---|---|---|
x1 | y1 | x2 | y2 |
|---|---|---|---|
0 100
1 30
1 50
Examples
| Global Caption | Region Captions | Region Bounding Boxes (x_min,y_min,x_max,y_max) |
|---|