Potential to track more points in live_demo.py?

I am curious about the potential to run live_demo.py with better GPU in order to track more points in real-time.

Have you ever done some testing to run it on cloud? If not, what would be the bottleneck? One thing I could think of is the streaming delay between local and cloud, but I am not sure whether it's a big problem.

Maybe switching & stacking better GPU on local would be more straightforward?

The high-level question would be: Have you thought about ways to scale the model for more points tracking?

Although it might be hard to answer without any experiments, it's always good to have some discussions in advance! Appreciate any comments and feedbacks!