Hi, @jyunlee! Thank you for sharing your code. I have some questions about object point cloud generation.
I understand that the input object point cloud was created by sampling 2048 points from the surface of an ARCTIC object mesh. However, I would like to know more about the specific process involved in generating the object point clouds from the ATCTIC dataset.
The ATCTIC dataset contains not only the object meshes but also the poses of the objects. My question is: how can I use this pose information to sample and generate the object point clouds?Specifically, how should I handle the case when the joint angles between the two parts of the object are not zero?
Additionally, could you explain how you filtered the necessary data from the original dataset?
Lastly, would it be possible to provide the processed dataset?
Hi, @jyunlee! Thank you for sharing your code. I have some questions about object point cloud generation.
I understand that the input object point cloud was created by sampling 2048 points from the surface of an ARCTIC object mesh. However, I would like to know more about the specific process involved in generating the object point clouds from the ATCTIC dataset.
The ATCTIC dataset contains not only the object meshes but also the poses of the objects. My question is: how can I use this pose information to sample and generate the object point clouds?Specifically, how should I handle the case when the joint angles between the two parts of the object are not zero?
Additionally, could you explain how you filtered the necessary data from the original dataset?
Lastly, would it be possible to provide the processed dataset?
Thank you!