tpofinder recognizes textured planar objects in images.
It first matches local features from the scene to the models, and then estimates homographies by random sample consensus.
tpofinder uses CMake as build system. OpenCV 2.4 is required.
git clone https://code.google.com/p/tpofinder/
cd tpofinder
mkdir build
cd build
cmake ..
make`
You can use your webcam to feed images to tpofinder. Some object models have already been included in the data directory: show one of these objects (or a printout) to the camera and see whether it is recognized.
tpofind --webcam
It is also possible to detect objects on images given as path names on the command-line:
tpofind --file some-image.jpg
If you have a directory full of images, you can pass them to tpofind via standard input:
cd tpofinder find some-folder -iname "*.jpg" -type f | tpofind
There are a couple of unit tests. Run them from the root of the source directory as follows:
utest
You can also specify further options to the Google test runner, see:
utest --help
The project was created on May 11, 2012.