search

zhaobinglei / REGNet_for_3D_Grasping

RGENet is a REgion-based Grasp Network for End-to-end Grasp Detection in Point Clouds. It aims at generating the optimal grasp of novel objects from partial noisy observations.
MIT License
56 stars 10 forks source link

About the dataset #1

Closed liuxi1234 closed 3 years ago

liuxi1234 commented 3 years ago

Can you give a detailed description of the production process of the dataset used?Thanks

zhaobinglei commented 3 years ago

The old dataset has only about 200 grasps in each scene. In the last week I will share a new dataset. The data generation involves 4 steps.(1) simulate scenes and render observed point cloud (2)construct collision-free grasps. (3) obtain grasp quality scores (4) compute point grasp confidence for each point in the observed point cloud. The code is based on the S4G. Best wishes, Binglei

liuxi1234 @.***> 于2021年4月29日周四 上午10:16写道：

Can you give a detailed description of the production process of the dataset used?Thanks

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/zhaobinglei/REGNet_for_3D_Grasping/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/AICEB6B4KTRZERYSR4EANUDTLC6PFANCNFSM43YKXIMA .

zhaobinglei commented 3 years ago

Same as the #3 The link to the new dataset is https://pan.baidu.com/s/1alwaKQZt0IGE11FFSClxOg . And the extraction code is x79a. (Maybe you can only use a Chinese proxy to access the link.)