search

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
656 stars 138 forks source link

some problem about the depth value #50

Closed chjXu closed 3 years ago

chjXu commented 3 years ago

Hi, i use the model to estimate the pose from a 1920x1080 pictures, the depth value of the joint is 230 in camera coordinate. So i want to ask the number of 230 means 230mm in the camera coordinate? And the X and Y is 15, -15.

what's means of the X and Y and depth value? Thanks.

Daniil-Osokin commented 3 years ago

Hi! X, Y, and Z are 3D joint coordinates in some coordinate space (camera, world, etc.). I believe the units are centimeters.

chjXu commented 3 years ago

This model has a big error, right?

Daniil-Osokin commented 3 years ago

It has reasonable error for monocular 3D pose estimation, you are free to check if it is useful for your needs. It is about 100 mm MPJPE (mean per joint position error) on CMU Panoptic subset.

Daniil-Osokin commented 3 years ago

Hope, it is clear now.