Save table prediction in cells format #2892
Conversation
| @@ -8,9 +8,11 @@ | |||
| # unstructured.documents.elements.Image | |||
| from PIL import Image as PILImage | |||
| from PIL import ImageSequence | |||
| from unstructured_inference.models.tables import cells_to_html | |||
There was a problem hiding this comment.
in unstructured we use a decorator on functions to note if this function requires additional extras to run. And unstructured_inference is one of those extras. As a result we don't import from unstructured_inference at module level normally but import only where it is needed and decorate the function with @requires_dependencies("unstructured_inference"). You can find examples in this file.
There was a problem hiding this comment.
thank you for the explanation!
|
|
||
| def test_simple_table_cell_parsing_from_table_transformer_when_missing_input(): | ||
| table_transformer_cell = {"row_nums": [], "column_nums": [], "cell text": "text"} | ||
| with pytest.raises(ValueError): |
There was a problem hiding this comment.
lets add a expected error message pattern matching here as well to make this test more explicit
|
|
||
| def test_simple_table_cell_parsing_from_table_transformer_when_missing_row_nums(): | ||
| cell = {"row_nums": [], "column_nums": [1], "cell text": "text"} | ||
| with pytest.raises(ValueError) as exception_info: |
There was a problem hiding this comment.
| with pytest.raises(ValueError) as exception_info: | |
| with pytest.raises(ValueError, match=f'Cell {str(cell)} has missing values under "row_nums" key'): |
| def test_simple_table_cell_parsing_from_table_transformer_when_expected_input(): | ||
| table_transformer_cell = {"row_nums": [3, 2, 1], "column_nums": [6, 7], "cell text": "text"} |
There was a problem hiding this comment.
lets use parametrize and add two more test cases:
- row nums has only 1 value
- column nums has only 1 value
| width = len(column_nums) | ||
| height = len(row_nums) | ||
|
|
||
| return cls(x=x, y=y, w=width, h=height, content=tatr_table_cell.get("cell text", "")) |
There was a problem hiding this comment.
we can combine 36-42 as one line
| width = len(column_nums) | |
| height = len(row_nums) | |
| return cls(x=x, y=y, w=width, h=height, content=tatr_table_cell.get("cell text", "")) | |
| width = len(column_nums) | |
| height = len(row_nums) | |
| return cls(x=min(column_nums), y=min(row_nums), w=len(column_nums), h=len(row_nums), content=tatr_table_cell.get("cell text", "")) |
| x: int | ||
| y: int | ||
| w: int | ||
| h: int |
There was a problem hiding this comment.
lets use full names here to make the code read more clear
There was a problem hiding this comment.
I have used 'w' and 'h' intentially to match deckard format as is currently used in mini-holistic. Then, for example, same code can be used during evaluation
plutasnyy
force-pushed
the
feat/table-as-cells
branch
from
3da2937
to
56a97f2
Compare
This pull request add metrics that are calculated based on table_as_cells instead of text_as_html. This change is required for comprehensive metrics calculation, as previously every colspan or rowspan predicted was considered to be an incorrect predicted (even if it was correct prediction) This change has to be merged after #2892 which introduces table_as_cells field
This pull request allows to return predictions in raw cell representation from table transformer. It will be later used to save prediction in a cells format for simpler metrics calculation.
This PR has to be merged, after Unstructured-IO/unstructured-inference#335