This package give a metric to evaluate table cell detection based on table cell bounding boxes.
You can install the package using pip:
pip install tabcell_score
Input format: We follow the format of the json label file from labelme.
{
"shapes":[
"label": "cell",
"points": [
[
0.,
0.,
],
[
13.,
13.
]
]
]
}
In code use:
from tabcell_score import TabCellScore
# Specify the label and image path
gt_path = r"sample\gt.json"
pred_path = r"sample\pred.json"
img_path = r"sample\image.jpg"
tabcellscore = TabCellScore(iou_threshold = 0.95)
# Output score only
score = tabcellscore(gt_path, pred_path)
# Output score and visualize on image and save to "result.jpg"
score = tabcellscore(gt_path, pred_path, img_path, "result.jpg")
# Output score and visualize on empty image and save to "result.jpg"
score = tabcellscore(gt_path, pred_path, None, "result.jpg")
print(score)
This project is licensed under the Apache License 2.0 - see the LICENSE file for details.
"Hope this package could help you in table cell evaluation." Phạm Phú Ngọc Trai (from akaOCR).