atenpas / gpd

Detect 6-DOF grasp poses in point clouds
BSD 2-Clause "Simplified" License
616 stars 233 forks source link

Is the training data trained by a complete point cloud or a partial point cloud? #137

Open qyp-robot opened 1 year ago

qyp-robot commented 1 year ago

Hi! @atenpas I find the object data from bigbird contain conpleted mesh module and partial point cloud, and the generate c++ code need 20 numbers point cloud module, when I use the parcial point cloud that view by singel camera, the antipodal graspa does not exist;I know need two singel camera point cloud connect ; or I can ues complete data to train? like this: ![Uploading image.png…]()

what's more, complete module is correspond to the parcial point cloud ,which is the groudtruth for it. What's the use?

qyp-robot commented 1 year ago

image

qyp-robot commented 1 year ago

@atenpas I'm sorry to bother you,How can I obtain a point cloud with 20 views of an object for trained? the picture as the follow: image