Comment by aantix

Don't these datasets already exist in terms of the real-world samplings we have when training robots to do every day tasks?