search

fabro66 / GAST-Net-3DPoseEstimation

A Graph Attention Spatio-temporal Convolutional Networks for 3D Human Pose Estimation in Video (GAST-Net)
MIT License
314 stars 69 forks source link

Human3.6M evaluation with non-GT 2D keypoints #39

Open guerrifrancesco opened 3 years ago

guerrifrancesco commented 3 years ago

Hi, I would evaluate GAST-Net using 2D keypoints from a 2d pose estimation framework. In the code you provided, 2D keypoints for evaluation are taken by the projection of the ground truth 3D keypoints. In your paper, I've seen that you've done this using CPN. Could you provide the code you've used? Thanks

fabro66 commented 3 years ago

Hi~

The 2D keypoints detected from CPN used in the article is from VideoPose3D.

We also provide a tutorial on how to run our model on custom videos. In this tutorial, we adopt HRNet as 2D pose detector. Please check it for more details.