@Yuliang-Liu
Hi Liu,
I'm trying to evaluate ABCNet on totaltext/CTW1500 with full lexicon. But I noticed that this repo didn't provide full-lexicon result and evaluation code on both datasets. I checked the totaltext repo and it said full lexicon consists of all words in test set.
So I'm wondering, is there an official script/code/rule for how to replace predicted words with lexicon words? And can you provide your script/code/rule, on both totaltext and CTW1500, if possible?
Thanks.
@Yuliang-Liu Hi Liu, I'm trying to evaluate ABCNet on totaltext/CTW1500 with full lexicon. But I noticed that this repo didn't provide full-lexicon result and evaluation code on both datasets. I checked the totaltext repo and it said full lexicon consists of all words in test set. So I'm wondering, is there an official script/code/rule for how to replace predicted words with lexicon words? And can you provide your script/code/rule, on both totaltext and CTW1500, if possible? Thanks.