It seems like their training is more geared to Known object types in front, and arbitrary objects on either side
It seems like their training is more geared to Known object types in front, and arbitrary objects on either side