[Docs] Why do the results of anbinet have only lowercase letters but not uppercase letters? The test results are all in green boxes, and some of the images are small, making the test results blurry.

open-mmlab / mmocr

OpenMMLab Text Detection, Recognition and Understanding Toolbox

https://mmocr.readthedocs.io/en/dev-1.x/

Apache License 2.0

4.27k stars 743 forks source link

[Docs] Why do the results of anbinet have only lowercase letters but not uppercase letters? The test results are all in green boxes, and some of the images are small, making the test results blurry. #1892

Open OuXiao514 opened 1 year ago

OuXiao514 commented 1 year ago

Branch

master branch https://mmocr.readthedocs.io/en/latest/

📚 The doc issue

Why do the results of anbinet have only lowercase letters but not uppercase letters? The test results are all in green boxes, and some of the images are small, making the test results blurry.

Suggest a potential alternative/fix

No response

Mountchicken commented 1 year ago

Hi @OuXiao514 Both the language decoder and vision decoder in ABINet are trained in lowercase, consequently, it can only output lowercase even though the text image is in uppercase.

OuXiao514 commented 1 year ago

What kinds of text recognition methods can be case-sensitive in mmocr?

Mountchicken commented 1 year ago

MASTER, SAR, RobustScanner, NRTR, SATRN, ASTER

OuXiao514 commented 1 year ago

I have another question, why are the real label and predicted results in green boxes when visualizing the model test results?