Open kxhaaa opened 10 months ago
Did you try it? How did it work out?
Hi I think that depends on what kind of sampling you are considering. If it is regular subsampling, e.g., decreasing subsampling from 30 to 10, that is probably okay. But if you are considering uneven sampling which may break the temporal coherence assumption that our method makes, then I think our method might have difficulty handling that.
Hi,
Congratualate for the amazing work!
I wonder if it is possible to sample some frames and train these data when I hope to track the whole frames of a video sequence. or I need to train all the frames of a long video and cost a lot of time.