search

eth-siplab / AvatarPoser

Official Code for ECCV 2022 paper "AvatarPoser: Articulated Full-Body Pose Tracking from Sparse Motion Sensing"
https://siplab.org/projects/AvatarPoser
MIT License
277 stars 46 forks source link

Positional Encoding? #6

Closed VimalMollyn closed 1 year ago

VimalMollyn commented 1 year ago

Do you apply any positional encoding to the input before feeding it to the transformer encoder?

jiaxi-jiang commented 1 year ago

Hi, we do not apply any positional encoding because a performance drop can be observed by doing that.