cs-chan / Total-Text-Dataset

Total Text Dataset. It consists of 1555 images with more than 3 different text orientations: Horizontal, Multi-Oriented, and Curved, one of a kind.
BSD 3-Clause "New" or "Revised" License
747 stars 142 forks source link

why do detection_filtering in the Deteval.py #11

Closed xiliu closed 5 years ago

xiliu commented 5 years ago

1,why do detection_filtering in the Deteval.py 2,why delete groundtruths which label is '#'

ckchng commented 5 years ago

Hi there,

It's to remove 'do not care' text region (which usually are languages that the dataset doesn't consider, illegible text regions, etc.). According to the standard protocol (like ICDAR2015 and COCO-Text), these text regions are not supposed to contribute to the evaluation.

xiliu commented 5 years ago

Thank you for replying, i get it.