Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Hand writing recognition for tamil #22

Open
tshrinivasan opened this issue Apr 9, 2019 · 7 comments

Comments

Projects
None yet
3 participants
@tshrinivasan
Copy link
Contributor

commented Apr 9, 2019

Hand writing recognition for malayalam
http://gitlab.com/santhoshtr/hand
https://thottingal.in/projects/hand/

make it for tamil.

Having ligature data in a file for example Tamil.json list like Malayalam.json. The data is very simple . it is x,y coordinates of a sample drawing of a letter. Just draw something on the pad, copy the console output to that json, rebuild. Next time onwards, those strokes will be recognized. This training step is documented in gitlab. This can be done with anybody who knows basics of programming and tamil script.

@tshrinivasan

This comment has been minimized.

Copy link
Contributor Author

commented Apr 9, 2019

@Kondasamy

This comment has been minimized.

Copy link

commented Apr 10, 2019

I will give a try on this project.

@tshrinivasan

This comment has been minimized.

Copy link
Contributor Author

commented Apr 10, 2019

@Kondasamy

This comment has been minimized.

Copy link

commented Apr 10, 2019

Please refer the project progress here - https://github.com/Kondasamy/tamil-handwriting-recognition

Goal:

  1. Add Tamil font coordinates for each ligatures and check the correctness
  2. Improve the user interface of the website (Dangling UI in Mobile)
  3. Improvise the algorithm to accurately detect characters (reduce false positives as much as possible)
@tshrinivasan

This comment has been minimized.

Copy link
Contributor Author

commented Apr 12, 2019

from @santhoshtr

I made some changes to code so that you can chose a language from UI. Added a small dummy training data for Tamil as well. You can see it at https://thottingal.in/projects/hand/
But if you try to add data, don't add lot of letters at once. Add a dozen and stop. This is because I am atill experimenting to see what level of data points granularity required for good performance and good accuracy. That is, I can define a circle using 4 points, 10 points, or 100 points. But this choice has impact on performance and accuracy.
Once we baseline this issue, we can go ahead for wider letter coverage.

@santhoshtr

This comment has been minimized.

Copy link

commented Apr 12, 2019

Thanks. You may not need a "port" of application, since it is mainly data driven. See how I integrated Tamil to application now. Feel free to fork and give pull requests, I can also give access to the repository.
But keep in mind about the data preparation. Go slow for few more days, while I experiment to see some optimal ways to define curve data.
When Jishnu tried to create android app, we realized that the main thread matching algorithm blocks the drawing UI as the data size increases. We might need to go with web worker for the matching part to resolve the issue.

Any ideas for improving is always welcome!

@Kondasamy

This comment has been minimized.

Copy link

commented Apr 12, 2019

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.