Amitabha Banerjee shows in a short video how a Vector recognizes another Vector without the need of custom markers.
This is based on the YOLO-framework (“you only look once”), a Python framework for object recognition. If you want to try this for yourself: you will need a CUDA-capable video card for this and you will need to train the system to recognize Vectors with multiple images of the robot.
And I am quite sure that this only runs through Vector and not on Vector, via the Python SDK, so technically it is not Vector that recognizes other Vectors, but the YOLO framework.
It’s a nice idea anyway. And once the learning process is finished it may be possible to transfer the learned image set to a robot without the need to redo the deep learning process. But that is an idea for the future if and when DDL releases anything substantial.