search

Masao-Taketani / FOTS_OCR

TensorFlow Implementation of FOTS, Fast Oriented Text Spotting with a Unified Network.
GNU General Public License v3.0
56 stars 15 forks source link

Recognition Error #6

Open JaeD-Shin opened 3 years ago

JaeD-Shin commented 3 years ago

1) We are applying an additional Korean language when studying on config.py in order to do Korean ocr. Does fosts support multi-language features?

2) Even though we completed the recognition test through the github readme, there seems to be a problem in detection, but we don’t know that is the cause of the error.

Masao-Taketani commented 3 years ago

Thank you for the comment. As for 1., yes, you can just add Korean characters to config.py. Then you can also train for Korean languages if I remember correctly. As for 2. what do you mean by "there seems to be a problem in detection"?

JaeD-Shin commented 3 years ago

Thank you for replying

As for 2. what do you mean by "there seems to be a problem in detection"? : Right now I'm learning through the GitHub readme. In the case of detection, the training proceeds well, and if the checkpoint created by recognition learning using the detection learning checkpoint is used in the test, the text box is not found.

Masao-Taketani commented 3 years ago

does the "text box" mean like the image below? image

JaeD-Shin commented 3 years ago

yes it is!

Masao-Taketani commented 3 years ago

Did you check outputs/ directory after executing the test.py?

JaeD-Shin commented 3 years ago

this is detcetion output.

detect1

But we did recognition training, the detection learning result with 2000 steps is not bad as follows de

but the recognition learning result cannot find any text box.

JaeD-Shin commented 3 years ago

Should we consider encoding by any chance?