Hi
I'm new to OpenIMAJ and wondering if it's suitable for realtime recognition cursive handwriting.. Any tips on where to start would be most appreciated.
Thanks
Amr
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
There isn't anything built-in, but I guess OpenIMAJ (coupled with it's dependencies) has all the key features you would need... What kind of recognition approach are you going to take - one based on pen stroke formation (i.e. tracking the coordinates of the pen as the user writes) or one based purely on the analysis of the previously written text in the image?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thanks Jonathon for your prompt reply..
I'm thinking to gather the list of strokes as the use is typing, and once the typing stops ( a sec of no new strokes ), to pass the list of strokes to a class that would return the actual text.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
OK, there are some features that might be of use: the OpenIMAJ Polygon class could be used to represent each stroke (the Polygon class allows representing polylines as "open" polygons), and polygons can be simplified to reduce the number of vertices (the Ramer-Douglas–Peucker algorithm is implemented by the reduceVertices() method of Polygon).
The simplified strokes could then be fed into something that encodes them as a feature (maybe encoding them as a chain code or similar?) to which a classifier could be applied...
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi
I'm new to OpenIMAJ and wondering if it's suitable for realtime recognition cursive handwriting.. Any tips on where to start would be most appreciated.
Thanks
Amr
There isn't anything built-in, but I guess OpenIMAJ (coupled with it's dependencies) has all the key features you would need... What kind of recognition approach are you going to take - one based on pen stroke formation (i.e. tracking the coordinates of the pen as the user writes) or one based purely on the analysis of the previously written text in the image?
Thanks Jonathon for your prompt reply..
I'm thinking to gather the list of strokes as the use is typing, and once the typing stops ( a sec of no new strokes ), to pass the list of strokes to a class that would return the actual text.
OK, there are some features that might be of use: the OpenIMAJ Polygon class could be used to represent each stroke (the Polygon class allows representing polylines as "open" polygons), and polygons can be simplified to reduce the number of vertices (the Ramer-Douglas–Peucker algorithm is implemented by the reduceVertices() method of Polygon).
The simplified strokes could then be fed into something that encodes them as a feature (maybe encoding them as a chain code or similar?) to which a classifier could be applied...
Most appreciated