Efficency of V2 and adding more points?

Hey,

First of all, great work. I really enjoyed reading the paper and going over the code, too. There are two things that I want to understand:

1. You mentioned that cotracker2 is more efficient than cotracker1. Could you help me understand where that efficiency is coming from? The one place I found a difference was in the transformer to compute the deltas. Looking closely at it, the transformers in v1 and v2 are based on space-time attention; v2 also uses some virtual tracks. So, since the tokens are increasing, they should be heavier on the memory. If true, then how is the new block more efficient? Am I missing something?

2. In my current use case. I want to add more points as the video progresses, but only in those regions where _no visible_ points are found. One way is to add more query points at certain intervals and remove the redundant points through post-processing, but this would be highly inefficient. Is there a smarter way of going about it?

Start by reading the referenced paper and comparing the CoTracker v1 and v2 transformer implementations to understand the claimed efficiency difference. Then inspect how point queries are added and tracked over time; the issue does not define an acceptance test or a clear completion condition for either question.

Written by the indexing model from the issue text.