OCR Application for automated information input, for this project, tested on StudentID cards
Includes 3 major parts:
_ID Card Extraction from raw image
_Words and Letters Extraction
_Classification
ID Card Extraction:
_Extract image corners (round corners, not common sharp corners)
_Perspective Warp
Words and Letters Extraction
_Words Extraction:
__Extract lines of information from pre-measured coordinates
__Extract words from lines of information
_Letters Extraction:
__Extract letters (with small symbols) from extracted words
Classification: Very simple deep learning networks, as pre-processing was done well
_Letters: 95%
_Numbers: 95%
Conclusion: Pretty naive project, still works pretty good with the right set up and equipment