search

lmb-freiburg / hand3d

Network estimating 3D Handpose from single color images
GNU General Public License v2.0
802 stars 252 forks source link

how fast is it? #13

Closed dedoogong closed 6 years ago

dedoogong commented 6 years ago

can I get any inference speed measurments? can it be poosible to run on TX2 in real time?

zimmerm commented 6 years ago

I can run the RGB -> 3D pipeline on an GTX 1080 Ti at 0.026sec per iteration, which equals ~ 38fps. Time for loading, writing or visualizing data is excluded. Unfortunately I can't provide the runtime on a TX2.

julianweisbord commented 6 years ago

Can you share the code that you used to achieve 38 fps? I copied run.py and made a few adjustments to work with opencv and my webcam but I am only getting 5 fps on a GTX 1080 TI.

vasanthhr commented 5 years ago

@zimmerm / @julianweisbord

I am running on G3.4xLarge instance(high performance GPU from AWS), but run.py execution time takes close to 6 seconds per image. Please look into this below issue and help me if i am doing anything wrong here.

Could you please help me on this issue, I am going though same issue now:

https://github.com/lmb-freiburg/hand3d/issues/25