search

Vegetebird / StridedTransformer-Pose3D

[TMM 2022] Exploiting Temporal Contexts with Strided Transformer for 3D Human Pose Estimation
MIT License
335 stars 37 forks source link

GPU usage when running demo code - low fps #15

Closed SonNguyen2510 closed 1 year ago

SonNguyen2510 commented 2 years ago

Dear author,

Thank you for your amazing work. I trying run your demo code with in-the-wild video and I got only about 15fps with 3D pose estimation (with HRnet and yolo detection) on my RTX3090 GPU. I checked gpu usage and it use only about 2GB of my memory. I want to ask that is there any setting that limmit my gpu usage and anyway to increase fps when running demo? I have install CUDA 11.x already. According to the paper, the method can achieve real-time application. Am I doing wrong some where? sorry, I'm new to this deeplearning field. Thank you very much for your time!

Vegetebird commented 2 years ago

In our demo code, we plot the joints, which is time-consuming. Without visualization, our model is real-time.

SonNguyen2510 commented 2 years ago

@Vegetebird I guess It is real-tine only when you inference 3D model alone right? When I try to use 2D pose in combination with 3D pose without visualization, it cannot achieve realtime

Vegetebird commented 1 year ago

Our 2D-to-3D pose lifting model is realtime. The 2D pose detector is HRNet, which is time-consuming. Or you can use some lightweight 2D pose detector, such as OpenPose.