Forums
Identify the handwritten correction notes made by the teacher, especially the messy check marks and incorrect correction notes.
Hello,
Thank you for contacting us.
Since you posted another question in a different thread, I edited your initial post to merge the two.
To guide you toward the best solution, we need to understand which features you would like to use:
Do you capture and render the strokes yourself or do you use an Editor?
What is the ink content you would like to recognize: it is a mix of text, math, shape, drawing? Which text language?
Sharing some ink of what is not correctly recognized (picture, strokes, recognition output) would help us investigate.
Best regards,
Gwenaëlle
15210276853
It's just about recognizing handwritten check marks (✓) and errors (✗), or combinations of text and these two symbols, path points used in handwriting, ones drawn by oneself and collected, not drawn using your SDK. How can this be solved? Currently, it's not accurately recognizing sloppy check marks and error symbols, nor can it recognize mixed text.
Thank you for your update.
Here is an example of ink written with the 4.5 Demo in a Raw Content part multiple latin languages:
The corresponding JIIX export result is:
{
"type": "Raw Content",
...
"elements": [ {
"id": "raw-content/471",
"type": "Text",
"lang": "en",
"label": "This exercise is good! ✓",
}, {
"id": "raw-content/475",
"type": "Text",
"lang": "en",
"label": "This exercise is wrong! ✗",
} ]
"id": "MainBlock",
"version": "3"
}
So please, share an ink example and its recognition output so that we can answer your question and better guide you.
Best regards,
Gwenaëlle
15210276853
For the precise identification of any path point trajectory, there may only be correct and incorrect batch trajectories, or there may also be trajectories with text nesting.
Hello, I'm from the Android SDK. I would like to ask about the accuracy of the SDK in identifying single check marks or handwritten error symbols. Since the recognition rate for many handwriting is not high, especially for the recognition of mixed text and symbols, are there any good solutions? The functional requirement here is related to the teacher's grading results and evaluations.