Questions on the evaluation code for "Task I - Text Localization"

[alwc]

According to the RCTW-17 paper on arXiv, it says,

A detection is matched to a groundtruth and marked as true
positive if 1) IoU is over 0.5 and 2) the groundtruth is not
matched to another detection. When multiple detections are
matched to the same groundtruth, we only pick the one with
the highest IoU.

In the code eval_task1.py I don't see the part when multiple detections are matched to the same groundtruth, it only pick the one with the highest IoU. Right now it seems the code only match criteria 1), but not 2).

[WeihongM]

same question, is the evaluate code wrong? @bgshih