No detection on small images #1057
Unanswered
underthesand
asked this question in
Q&A
Replies: 1 comment 1 reply
-
|
Hi @underthesand 👋, another option would be to pad small images if you want to use the whole pipeline. |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
FYI @underthesand this is a workaround, not the longterm solution we'll work on 👍 |
Beta Was this translation helpful? Give feedback.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
On small images there is no detection but if I zoom out there is one.
how i can solve this problem please ?
Édit : As DocTR hasnt been trained on small chucks of texts the solutions are custom training or only using the recognition model if we already know text position.
Source #639
Big
Small
Beta Was this translation helpful? Give feedback.
All reactions