Update README.md
Browse files
README.md
CHANGED
|
@@ -77,8 +77,8 @@ Each example in the dataset contains the following fields:
|
|
| 77 |
- **`answer_idx`**: Index of the correct answer in the choices list (integer)
|
| 78 |
- **`datatype`**: Type of spatial reasoning task (value: "perspective")
|
| 79 |
- **`house_ind`**: House/scene identifier (integer)
|
| 80 |
-
- **`cam_position`**: Camera position coordinates as 3D
|
| 81 |
-
- **`cam_rotation`**: Camera rotation
|
| 82 |
- **`image_reason`**: Rendered image from the new perspective that the question is asking about. This provides the ground truth visualization showing what the scene looks like from the target viewpoint.
|
| 83 |
|
| 84 |
### Example
|
|
|
|
| 77 |
- **`answer_idx`**: Index of the correct answer in the choices list (integer)
|
| 78 |
- **`datatype`**: Type of spatial reasoning task (value: "perspective")
|
| 79 |
- **`house_ind`**: House/scene identifier (integer)
|
| 80 |
+
- **`cam_position`**: Camera position coordinates of first image and second image as (x,y,z) 3D coordinate.
|
| 81 |
+
- **`cam_rotation`**: Camera rotation for the first image and second image. The rotation in THOR is represented as (x,y,z). x and z are always 0 for us. Hence, the two values represent the y values.
|
| 82 |
- **`image_reason`**: Rendered image from the new perspective that the question is asking about. This provides the ground truth visualization showing what the scene looks like from the target viewpoint.
|
| 83 |
|
| 84 |
### Example
|