reset_index

Dataset.reset_index(start_image_id: int = 0, start_annotations_id: int = 0, sort_images_by: None | str | Sequence[str] = 'relative_path', sort_annotations_by: None | str | Sequence[str] = ('image_id', 'category_id', 'box_x_min', 'box_y_min', 'box_width', 'box_height')) → Self

Reset index of self.images dataframe, and reset index of self.annotations However, keep the ‘image_id’ column in self.annotations pointing to the right rows in the self.images dataframe.

Note

Both images and annotations dataframes will be reorder according to specific columns. You can change them with the sort_images_by and sort_annotations_by parameters, but the default behaviour is:

• images dataframe will be reordered according to relative_path

• annotations dataframe will be reordered according to image_id, category_id and the bounding box coordinates, i.e. box_x_min, box_y_min, box_width and box_height

Parameters:

• start_image_id – Number at which the image index starts. This is used to construct two datasets without overlapping ids.

• start_annotations_id – Similar to start_image_id, number at which the annotations index starts.

• sort_images_by – columns to sort the images dataframe by. It is advised to chose a collection of columns that makes the sorting unique. If set to None or an empty sequence, will no sort the images dataframe before applying a range index to it. Defaults to relative_path

• sort_annotations_by – columns to sort the annotations dataframe by. It is advised to chose a collection of columns that makes the sorting unique. If set to None or an empty sequence, will not sort the annotations dataframe before applying a range index to it. Defaults to ("image_id", "category_id", "box_x_min", "box_y_min", "box_width", box_height").

Returns:

Dataset with self.images and self.annotations with updated indexes

See also

• related tutorial

• reset_index_from_mapping()

• match_index()

Example

>>> from lours.utils.doc_utils import dummy_dataset
>>> example = dummy_dataset(10, 10, seed=2)
>>> example.iloc[1::2]
Dataset object containing 5 images and 4 objects
Name :
    argue_be_structure
Images root :
    what/way
Images :
    width  height          relative_path   type  split
id
1     472     892        also/policy.gif   .gif  train
3     506     602      increase/pull.jpg   .jpg    val
5     401     281         would/off.jpeg  .jpeg  train
7     831     375        ahead/truth.bmp   .bmp  train
9     993     334  husband/whatever.jpeg  .jpeg   eval
Annotations :
    image_id  category_str  category_id  ...   box_y_min   box_width  box_height
id                                       ...
0          1  relationship            3  ...  606.391824   29.194750  194.387036
1          7  relationship            3  ...  313.193702  230.609055    5.269920
5          9        simply           25  ...  198.210135  474.192703   57.594892
9          9         table            7  ...   60.522880  425.022919  144.458578

[4 rows x 8 columns]
Label map :
{3: 'relationship', 7: 'table', 25: 'simply'}
>>> example.iloc[1::2].reset_index(10, 5)
Dataset object containing 5 images and 4 objects
Name :
    argue_be_structure
Images root :
    what/way
Images :
    width  height          relative_path   type  split
id
10    831     375        ahead/truth.bmp   .bmp  train
11    472     892        also/policy.gif   .gif  train
12    993     334  husband/whatever.jpeg  .jpeg   eval
13    506     602      increase/pull.jpg   .jpg    val
14    401     281         would/off.jpeg  .jpeg  train
Annotations :
    image_id  category_str  category_id  ...   box_y_min   box_width  box_height
id                                       ...
5         10  relationship            3  ...  313.193702  230.609055    5.269920
6         11  relationship            3  ...  606.391824   29.194750  194.387036
7         12         table            7  ...   60.522880  425.022919  144.458578
8         12        simply           25  ...  198.210135  474.192703   57.594892

[4 rows x 8 columns]
Label map :
{3: 'relationship', 7: 'table', 25: 'simply'}