How to improve live pose tracking ? FPS Speed ? - Githubissues

Daniil-Osokin / lightweight-human-pose-estimation-3d-demo.pytorch

Real-time 3D multi-person pose estimation demo in PyTorch. OpenVINO backend can be used for fast inference on CPU.

Apache License 2.0

653 stars 137 forks source link

How to improve live pose tracking ? FPS Speed ? #62

Closed pixellib closed 3 years ago

pixellib commented 3 years ago

How to improve live pose tracking ? FPS Speed ?

Daniil-Osokin commented 3 years ago

Hi! If you want to run the demo faster, you can infer the network on a smaller resolution, just pass --height-size 192 option to the script.

pixellib commented 3 years ago

Thank you... i will check...

pixellib commented 3 years ago

thank you @Daniil-Osokin its working, if we reduce size the tracking accuracy less ? or not ?

Daniil-Osokin commented 3 years ago

Hi! It depends on the data you are testing on. If persons are quite big in an input image, then results should be ok.

Daniil-Osokin commented 3 years ago

Hope, it helped.