brown-ivl / manus

[CVPR 2024] MANUS: Markerless Grasp Capture using Articulated 3D Gaussians
https://ivl.cs.brown.edu/research/manus.html
15 stars 6 forks source link

How to use raw videos as input #5

Closed 1524101973 closed 3 months ago

1524101973 commented 3 months ago

Hi , Thanks for sharing the awesome work! I've noticed the code seems to be designed for pre-processed data. Is there a way to use raw videos as input? If so, could you provide some guidance or point me to relevant resources? Any help would be greatly appreciated!

coreqode commented 3 months ago

Hi, raw videos were used only to get the hand poses. However, they can be used as a dataset for multiple downstream problems when combined with pose annotations.