search

CMU-Perceptual-Computing-Lab / openpose

OpenPose: Real-time multi-person keypoint detection library for body, face, hands, and foot estimation
https://cmu-perceptual-computing-lab.github.io/openpose
Other
31.09k stars 7.85k forks source link

How to use Model BODY_25B with Python API for better detection #1930

Open Christoph9402 opened 3 years ago

Christoph9402 commented 3 years ago

Issue Summary

Hello, I want to use the model Body_25B since the accuracy of Model Body_25 was not as good as axpected. I downloaded both files and have the correct folder structure as described in the documentation. However, I want to use the python API. Do you know how to activate the model there? The documentation says: "Run the official OpenPose with your usual flags, while adding the --model_pose BODY_25B. Note that this model has been trained for the original and default --net_resolution -1x368, so this flag is not required." I am not sure what to do when using Python. How do I have to modify my script? Below are 2 images (1920x 540) where I was hoping to get (better) detections: Frame_undistorted140 Frame_undistorted220 Are my expectiations too high, or is it possible that the model Body_25B could help? I am grateful for any other recommendations. I want to extract Keypoints from the body and feet. Face and Hands are not needed.

Type of Issue

Your System Configuration

  1. OpenPose version: Latest Version from Github

  2. General configuration:

    • Installation mode: CMake (Windows)
    • Operating system: Windows
    • Operating system version : Windows 10
    • Release or Debug mode? (by default: release): release
    • Compiler: Visual Studio 2019 (Windows)

  3. Non-default settings:

    • 3-D Reconstruction module added? (by default: no): no
    • Any other custom CMake configuration with respect to the default version?: Python API enabled

  4. 3rd-party software:

    • Caffe version: Default from OpenPose
    • CMake version 3.20.1
    • OpenCV version: 4.0.1

  5. If GPU mode issue:

    • CUDA version : 11.1
    • cuDNN version: 8.1.0
    • GPU model : RTX 3070

  6. If CPU-only mode issue:

    • CPU brand & model: Ryzen 3600
    • Total RAM memory available: 16Gb

  7. If Python API:

    • Python version: 3.8.8
    • Numpy version: 1.19.2

  8. If Windows system:

    • Portable demo or compiled library? compiled libary
stale[bot] commented 3 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

ovladuk commented 2 years ago

did you get it working? i can't believe no one has replied