VAD
VAD copied to clipboard
Perception without tracking task
Why does Agent Query directly perform motion transformer without performing tracking tasks? Will this reduce performance?