OCR stands for Optical Character Recognition software. As part of a student project, our objective is to make a program capable of extracting text from images. It must be written in the C language, rely on a neural network and be used through a GUI. To learn more, please take a look at the book of specifications.
Dependencies: SDL2, SDL2_image, GTK+ 3 and Hunspell.
- Clone this repository with
git clone git@github.com:NoneOfAllOfTheAbove/OCR.git. - Compile the project by running the command
makein the project folder. - Execute the program with
./bin/OCR.
Currently implemented:
- Advanced preprocessing (efficient binarization, noise canceling, contrast enhancement)
- Detect paragraphs, lines, words and characters
- A pretrained neural network to recognize characters
- Simple GUI to load an image and export its extracted text
- Postprocessing step (spell check)
Features we are working on:
- De-skew
- Improve segmentation (export as HTML)
- Retrain the neural network
Refer to CONTRIBUTING.md.