______ ______ _______
/ \ / \ | \
| $$$$$$\| $$$$$$\| $$$$$$$\
| $$ | $$| $$ \$$| $$__| $$
| $$ | $$| $$ | $$ $$
| $$ | $$| $$ __ | $$$$$$$\
| $$__/ $$| $$__/ \| $$ | $$
\$$ $$ \$$ $$| $$ | $$
\$$$$$$ \$$$$$$ \$$ \$$
BY SOYCENT
README
OCR is a tool for converting images to text using neural networks.
Currently, the tool is capable of loading an image in its system,
segmentate the lines and characters and pass it to the neural network.
It is also possible to train this neural network.
USAGE
./main <path to image>
Converts the specified image to a text. Result is printed
on stdout. <path to image> must be a valid path to an image
containing text. The recommended image format is JPEG.
Training:
./main --train <training set folder>
Trains the neural network using the provided training sets and
saves the result. <training set folder> must be a valid path to
a directory containing valid training images set. The given set
is available in the trainset folder.
Example: ./main --train trainset
A valid training image contains the following caracters in order:
abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ,.
It is recommended to use a high letter spacing to avoid any issue
during the segmentation.
To use the GTK interface, launch with:
./main_gtk
COMPILING
To compile the GTK interface, use the following:
make ui
To compile the CLI version, use the following:
make cli
To compile both, just use:
make
CURRENT STATUS
The OCR tool project reached the end. It still contains bugs.