This is a proposal for extending and improving the OpenCV text module. In brief I propose implementing word-level classification with deep CNN for text recognition, as this strategy has proven superior to character level. I propose to implement an optional subset for training models introducing external dependencies and versatile, optionally light weight, feed-forward recogniser that can be used wherever OpenCV is deployable.

Organization

Student

Anguelos

Mentors

  • Lluis Gomez
close

2016