Datasets:
ArXiv:
License:
Update README.md
Browse files
README.md
CHANGED
|
@@ -8,6 +8,8 @@ This document describes the acquisition of the pre-training data used by OS-ATLA
|
|
| 8 |
**Notes:** In GUI grounding data, the position of the target element is recorded in the `bbox` key, represented by `[left, top, right, bottom]`.
|
| 9 |
Each value is a [0, 1] decimal number indicating the ratio of the corresponding position to the width or height of the image.
|
| 10 |
|
|
|
|
|
|
|
| 11 |
The data we released is divided into three domains: mobile, desktop and web.
|
| 12 |
|
| 13 |
All annotation data is stored in JSON format and each sample contains:
|
|
|
|
| 8 |
**Notes:** In GUI grounding data, the position of the target element is recorded in the `bbox` key, represented by `[left, top, right, bottom]`.
|
| 9 |
Each value is a [0, 1] decimal number indicating the ratio of the corresponding position to the width or height of the image.
|
| 10 |
|
| 11 |
+
The data stored in this dataset consists of raw data containing only element grounding information. When training a model, you need to use the corresponding prompts to process these data.
|
| 12 |
+
|
| 13 |
The data we released is divided into three domains: mobile, desktop and web.
|
| 14 |
|
| 15 |
All annotation data is stored in JSON format and each sample contains:
|