WHAT |
Entirely custom implementation of OCR to extract spreadsheets |
WHEN |
March-April 2014 |
WHO |
Me, Pavel Komarov, Ricky Liou, Hank McCord, Miguel Serrano |
WHERE |
Georgia Tech, ECE 4580 |
WHY |
I had to write many tables in lab and manually transfer those to my computer, which struck me as silly |
HOW |
Liking that diagram? Yes, we implemented all of that, even the custom OCR - that used Extreme Neural Nets that took much too long to debug. In the end, we did get a whole working pipeline (as yuo can see in the images below), and so I was fairly happy with the results. Now, just need to reimplement all of it in Java and port it to Android… |
LINKS |
|
PICS |
|
VIDS |
|