Dear organizers,
It is my understanding that the evaluation pipeline of this challenge is:
1) generate images from each test sequence;
2) run the three given text recognition algorithms on the generated images;
3) report the accuracy of the three text recognition algorithms.
My question is, are we allowed to utilize the three given text recognition algorithms in step 1)? A straightforward example is to run text recognition on each frame and select frames with high confidence as keyframes. Could you please specify whether such processing is allowed?
Posted by: laod @ April 4, 2023, 1:13 a.m.