search

eth-ait / dip18

Deep Inertial Poser: Learning to Reconstruct Human Pose from Sparse Inertial Measurements in Real Time
GNU General Public License v3.0
146 stars 47 forks source link

Can sensor data be directly input to the network without rotating the axis? #14

Closed NINGgege111 closed 3 years ago

NINGgege111 commented 3 years ago

Thank you for your work. I am experimenting with another sensor, but the incoming axis is always different from the unity game. Can you help me solve this problem?

YinghaoHuang91 commented 3 years ago

Hi the model is trained with rotation and acceleration normalized wrt root joint framewise, so to test it on your own data you also need to do the same normalization for the network to work reasonably. For the details please refer to the Section 4.3 of the paper.