Imago Go Image Recognition
There is also an experimantal Haskell version of Imago.
- Python 2.7 (including dev)
- PIL >= 3.0
- matplotlib (for generating debug images)
- openCV (if you want to use a camera, otherwise not needed)
- for documentation:
make in this directory.
./imago image.jpg to extract game position from image.jpg.
./imago -m image.jpg to manually select grid position.
./imago image000.jpg image001.jpg image002.jpg ... to produce a game record from a sequence of images, one for every move. Use
-S option to select SGF output.
./imago --help for help and list of all options.
The imago test suite lives in the
imago-tests repository, which is a
submodule of this repository.
git submodule init; git submodule update (or clone this repository with
git clone --recursive) to get the test data.
To run the tests, build imago and run
./runtests from the root of the
The runtests program runs imago on each input image and compares the output to the expected. Any discrepancy is a test failure.
Failing tests can be disabled by moving them into a directory named
"skip". "skip" directories are by default ignored by the runtests
./runtests --all to include tests marked "skip".
- Tomáš Musil (@tomasmcz)
- Sebastian Kuzminsky (@SebKuzminsky)
- Nicolas Rochette