How can I speed up the inference when testing on large-scale videos?

vye16 / slahmr

MIT License

441 stars 50 forks source link

How can I speed up the inference when testing on large-scale videos? #23

Closed linjing7 closed 1 year ago

linjing7 commented 1 year ago

Hi, thanks for your excellent work! I have a question regarding speeding up inference when testing on large-scale videos. In README, You recommend preprocessing in advance. However, the tracking is still run with batch_size=1, which is quite slow. Is it possible to use a larger batch_size to speed up the tracking? BTW, could you please tell me the advantage of preprocessing in advance?

smandava98 commented 12 months ago

Hi! Did you manage to figure this out? I am having the same issue :/