First of all, great work!
According to the Sam3 object tracking doc, to track objects in a video:
- set video
- generate masks using text prompt
- remove unwanted object by id
- propagate
The problem is that the text prompt cannot segment the exact objects, so I want to use a different model to get the segmentations, so the steps can be:
- get segmentation using other model
- set video and the segmentions of the first frame
- propagate
Basically just use the tracking ability of Sam3. Can this be done? Thanks in advance.