search

BII-wushuang / Lie-Group-Motion-Prediction

Code for CVPR2019 Towards Natural and Accurate Future Motion Prediction of Humans and Animals
https://coderstellaj.github.io/Hierarchical-Motion-Recurrent-Network-Website/
MIT License
97 stars 18 forks source link

Data preprocessing #11

Open SudoAniki opened 1 year ago

SudoAniki commented 1 year ago

Hello! I was wondering how the preprocessing of the human lie datatype was made, specifically the conversion from the xyz joints to the lie group. Following your matlab steps, I ended up with a lie output of (nframes njoints 6), but the preprocessed dataset you provided in data/h3.6m is formated in something like (mframes * 99) in exponential map format as show in asheshjain399's Structural RNN.

I'm new in this whole lie group theory, so how did you convert your output of lie parameters (nframes njoints 6) to the SRNN's format in exponential map (mframes * 99)? Any help would be greatly apreciated.

phaiyen0493 commented 1 year ago

same. I hope the author can explain, or provide the code to convert the lie parameters (nframes njoints 6) too. I stil cannot get how the exponential map turned to out be 99 parameters.