An open-source project dedicated to tracking and segmenting any objects in videos, either automatically or interactively. The primary algorithms utilized include the Segment Anything Model (SAM) for key-frame segmentation and Associating Objects with Transformers (AOT) for efficient tracking and propagation purposes.
Hi there, thanks for the code, it's been super cool to see and use!
I'm using a set of videos (think self-driving) where new objects enter and leave the frame all the time, so I want to be able to delete objects that leave the frame fast from tracker memory (or flush objects very aggressively from the tracker if they haven't been tracked for a short period). That way I can avoid the issue of false positives where new objects are identified as previous objects due to long-term memory even though they shouldn't be.
Is there a way to tune DeAOT towards this "aggressive removal and recreation"?
Hi there, thanks for the code, it's been super cool to see and use!
I'm using a set of videos (think self-driving) where new objects enter and leave the frame all the time, so I want to be able to delete objects that leave the frame fast from tracker memory (or flush objects very aggressively from the tracker if they haven't been tracked for a short period). That way I can avoid the issue of false positives where new objects are identified as previous objects due to long-term memory even though they shouldn't be.
Is there a way to tune DeAOT towards this "aggressive removal and recreation"?