And additionally, will you be comparing model to unknown data, for example stamps of cars / birds and so on, for which we don't have data?
(Asking because this what happened in first competition, with check data drastically different from training data)
The dimensions are abstract - the x-axis is the 'slow time', the y-axis is the 'fast time' and I/Q represent the real and imaginary parts of the received signal. Participants can still use various processing techniques based on those abstract dimensions.
The models will be tested on segments of humans or animals as in the training data. Note that one of the main goals of this competition is generalization. The models will be tested, as in "real life", on segments from new geolocations, dates, etc.
(MAFAT Challenge Team)