facebookresearch / co-tracker

CoTracker is a model for tracking any point (pixel) on a video.
https://co-tracker.github.io/
Other
3.75k stars 246 forks source link

Cotracker3 Inference time #110

Open jun297 opened 2 weeks ago

jun297 commented 2 weeks ago

Hi, thank you for sharing the nice work!

I just read the CoTracker3 paper and I wondered how fast CoTracker3 compared to Cotracker1 or 2.

What I only know is the 'Time' column in RoboTAP evaluation result Then can I assume that Cotracker3 is 2 times faster than Cotracker 1 or 2? or can I check the detailed information regarding inference time? image

zhw-zhang commented 2 weeks ago

Thank you for sharing the great work, I am also interested in the inference time. I would like to know if the time required for a 16-frame video at 512x512 resolution is calculated as 200 microseconds 512 512 * 16, which would be approximately 838 seconds, or roughly 13 minutes?