search

ethz-asl / vgn

Real-time 6 DOF grasp detection in clutter.
BSD 3-Clause "New" or "Revised" License
257 stars 55 forks source link

tsdf or pc? #14

Closed wd117 closed 2 years ago

wd117 commented 3 years ago

First i must admit that you have shared a great work, i'm really interested in this so i have read your paper. In your paper,you said: While point clouds are able to identify object surfaces in space, we believe that the additional distance-to-surface information provided by a TSDF representation can improve overall grasp detection performance I can't get it, why do you think tsdf is better than pc? Hope to get your help,thanks!

mbreyer commented 2 years ago

Hi wd117, the idea was that a TSDF not only encodes information about the object surface itself (like a point cloud), but also whether voxels lie outside or inside an observed surface which could be advantageous for learning grasps. It would be interesting to do a direct comparison to point clouds though.

wd117 commented 2 years ago

Thanks!