Total Text Does not give character level annotation

[anmol4210]

No description provided.

[anmol4210]

my bad

[anmol4210]

Hi,
Total text gives character level mask but it does not gives coordinates of every character in the image. But you have mentioned character level annotation as yes.

[ZumingHuang]

Thank you for your comment.

Since Total-text dataset provides character-level mask, we can use connected component analysis (CCA) to obtain the coordinates of every character based on the character-level mask.

By the way, I think it would be more appropriate to say pixel-level annotation as yes.