|
|
--- |
|
|
task_categories: |
|
|
- image-to-image |
|
|
size_categories: |
|
|
- 1K<n<10K |
|
|
tags: |
|
|
- image-inpainting |
|
|
- object-removal |
|
|
pretty_name: 'Dataset Inpainting AI4LIFE ' |
|
|
--- |
|
|
|
|
|
# Custom Object Removal Dataset |
|
|
|
|
|
This dataset is designed for **object removal** and **image inpainting** tasks. |
|
|
|
|
|
Each sample includes: |
|
|
- An object image (`object/` folder) |
|
|
- A background-only image (`background/` folder) |
|
|
- A binary object mask (`mask/` folder) |
|
|
- A bounding box annotation file (`annotations_bbox.json`) |
|
|
|
|
|
Images and masks are matched by filename (`1.jpg`, `2.jpg`, `3.jpg`, ...). |
|
|
|
|
|
--- |
|
|
## Folder Structure |
|
|
background/ 1.jpg 2.jpg ... object/ 1.jpg 2.jpg ... mask/ 1.jpg 2.jpg ... annotations_bbox.json |
|
|
|
|
|
- `background/`: Images of the scene after object removal. |
|
|
- `object/`: Images containing the target object. |
|
|
- `mask/`: Binary masks (white = object, black = background). |
|
|
- `annotations_bbox.json`: Bounding box annotations in the format `{image_id: [x_min, y_min, width, height]}`. |
|
|
|
|
|
--- |
