support a sentence as text_prompt for Ground and Tracking?

[shixishi]

For custom video input, does the supported text prompt have to be a word (i.e. a word representing a certain category), can it be a sentence?

[rentainhe]

Sure, I think it can be achieved by both Grounding DINO and Florence-2, we will consider support it using Florence-2 in the future release. But we only ground phrases in one specific frame and track the object in the following frame, if you need more complex applications, you may modify it by yourself.