search

clovaai / deep-text-recognition-benchmark

Text recognition (optical character recognition) with deep learning methods, ICCV 2019
Apache License 2.0
3.75k stars 1.1k forks source link

how to deal with text recognition with images whose width are more than 280 #120

Open jake221 opened 4 years ago

jake221 commented 4 years ago

thanks for your sharing, great work. I am using your code with CRAFT to deal with a chinese and english text recognition task. I set the image width as 280. And the width of training data of chinese is all 280, but the training data of english is like yours as MJ. However, the width of the output chinese text image of CRAFT is usually more than 280, and the english text image of Craft is less than 280. So the recognition accuracy for the real chinese document is very low. So can I solve this problem?

jake221 commented 4 years ago

image