Yeah, I've noticed this in the US as well where Autopilot gets very conservative about positioning of large trucks when viewed from non-main cameras. I believe this is one of the primary reasons for adding the birds-eye-view network to learn how to better place objects consistently with the main camera by taking the individual camera predictions and outputting a unified view.I know it's not FSD beta, but I wonder if that is an explanation for why when overtaking lorries (trucks) in inside lane of a UK motorway, on passing, the lorry then jumps into middle lane?
So pre-FSD beta, Autopilot is taking individual camera neural network predictions and manually calculating their positions and combining them, and this regularly causes duplicate or jumpy objects especially when the objects transition between cameras.