Generic Comments

ORK is able to run different object recognition pipelines concurrently while using a common infrastructure: data will come in through a Source that will then go to a Pipeline which will output information to a Sink.


Those are usually written in C++ and can come from any library out there. To be usable, they need to be wrapper in a C++ ecto cell as ORK is running everything in a big ecto graph. To have this cell foundable and easier to use, it also needs to be wrapped in a Python object (there is a base class for any of the types).

Unit Tests

To make sure your Source/Pipeline/Sink is compatible with ORK, we provide CMake macros to test your code. Use them as follows:

# for a detection pipeline
# for a training pipeline
# for a sink
# for a source
# for a generic Python script you want to test --help on