hkchengrex
commented
1 month ago On a high level, we have something like this: https://github.com/hkchengrex/Tracking-Anything-with-DEVA which is quite a complicated pipeline supporting new objects and different base models.
For something simpler like detection at the first frame + propagation, you can start with the scripting demo (by replacing mask with your automatic generation).
Hello, does the code support automatic mask extraction without human interaction, or how can I make simple changes to the code to extract mask based on prompt such as text prompt.