I have a simple strategy (pseudo-code) that details 11 novel arguments to extract 4 topological signals in handwritten “stroke networks” (i.e. the splits, merges, termination signals, etc. in any scanned handwritten document) that can be sequenced offline to facilitate the “non-temporal” (off-line) machine recognition of handwriting. The entire product is obviously much more complicated but I have proven test pages at each of the five main phases of this project. I need some help with integrating the strategy into a final product.
The US Post Office can scan handwritten zip codes, but their lexicon library is limited to just ten digits (0-9). I've talked to Dr. Srihari from SUNY, Buffalo, whose department was awarded the Postal grant to develop ICR for the US Mail service, but he said that the universe of family names in archive documents is too difficult at this time.
There are several billion pages of tax records in the world's archives, meticulously recorded by various middle class groups down through history to "tax themselves", but current access to any of these documents is through tedious "manual" searches. The Mormon Church has tried to retro-convert several big projects with the US Census records, but these usually involve manual "double-key" entry strategies -- not machine ICR digitization. They simply pay low-cost (but highly educated workers (usually in the Philippines) to type the same documents on two different machines, and then correct the discrepancies by matching the output.
I think that I can take this technology to a new level. I estimate that it will take about five weeks @ 6 hrs/ week for about 30 hours total for me to transfer the logic arguments (pseudo-code) to your developers.
Test Document (input):
Scanned microfilm pages from New York State Archives 1789 Tax Lists, -- (or any other scanned "handwritten" document) -- see attachment for sample input-output.
I'm available to meet Mon, Tu, Th, Sat, Sun @ any time.
Let me know what you think,
Hello! I am really interested in your project because I am skilled and experienced in machine learning and image processing programming. Can we please discuss further? Thank you!
Hi there! Thanks for posting the job. I am fully skilled and experienced in image processing and character recognition so I can help you with this project. I hope to discuss the project further with you. Thank you!