OCR Model. What's "anchor-words", and how to define them?

[dio_lee]

I build a OCR Model to get the code(商品コード)] from photo. But got the warning below.

“A warning was produced while training OCR Model ‘Read Code 2’: No anchor-words were found; defaulting to using top-left corner and document size for alignment.”

My question is how to make my anchor words(I think it’s “商品コード”) to be found.

[kg_fonkey]

Same as me.

[Chee]

Having the same problem. Anyone found out what this “anchor words” are?

[sodaa]

Any luck yet people? I have also been facing that issue all week

[Steve]

OCR is designed to be used with well-structured images, such as pictures of forms. It is not designed to work with arbitrary images.

An "anchor word" would be a word or words that appear in every image (such as the questions on a survey form) that would help the OCR recognize what form it's seeing and where in the form it's at.