search

michalfaber / tensorflow_Realtime_Multi-Person_Pose_Estimation

Multi-Person Pose Estimation project for Tensorflow 2.0 with a small and fast model based on MobilenetV3
Other
215 stars 65 forks source link

Inference Time #23

Closed AlexandrineRibeiro closed 3 years ago

AlexandrineRibeiro commented 3 years ago

Hi,

Thank you for the great job you are doing with this project.

I was testing this project using the python demo_video.py ... examples.

The processing time is around 0.15s per image using a Nvidia GTX1070 GPU.

As this repo is an upgrade to an older version to become compatible with the latest and recommended version of Tensorflow 2.0, I was expecting it to be faster.

Is it supposed to take so long?

AlexandrineRibeiro commented 3 years ago

Hi @michalfaber,

I noticed that the problem was due to the fact that the settings were being loaded into the loop (this line cfg = get_default_configuration()). I fixed the problem, now the settings are loaded only once at the beginning of the code.

I made a pull request to fix this in this repository. I hope you will consider my contribution.

Thanks.