search

facebookresearch / segment-anything-2

The repository provides code for running inference with the Meta Segment Anything Model 2 (SAM 2), links for downloading the trained model checkpoints, and example notebooks that show how to use the model.
Apache License 2.0
10.63k stars 848 forks source link

How to track just a few keypoints in the video? #163

Open gigasurgeon opened 1 month ago

gigasurgeon commented 1 month ago

I want to track single points on the video. I don't want to track the entire object, just a point. Is it possible to achieve this?

starphantom666 commented 1 month ago

same question

heyoeyo commented 4 weeks ago

There is an older paper called Tracking Everything Everywhere All at Once that seems like it would out-perform SAM for tracking points. If you wanted to use SAM, the best option is probably to start the video predictor with a mask that is very small around the point of interest.

gigasurgeon commented 4 weeks ago

@heyoeyo Thanks. I also found co-tracker by meta which also seems to outperform sam2 on tracking tasks