pytorch_6dof-graspnet
pytorch_6dof-graspnet copied to clipboard
jsll
bad results
hello,
Thank you very much for sharing this great work.
I have trained a VAE model following the instruction,it seems work well on the NPY dataset you given
But if I use the realsense d435i get the pointcloud data,the results are not ideal
Obviously i have made some mistakes,but i don't know where it happens,Where do you think i'm most likely to make mistakes,hope to get your help ,thank you very much.
One idea, have you normalized the point-cloud from the realsense? That is, have you subtracted the mean of the input point-cloud?
One idea, have you normalized the point-cloud from the realsense? That is, have you subtracted the mean of the input point-cloud?
I find the grasps are generated by the function ' estimator.generate_and_refine_grasps() ',and the only input is the pointcloud. In the function,the first step is 'prepare pc',i think during this the pc should have been normalized.
You are correct. Been a while since I used this code base so I have forgotten some details.
Maybe the problem is that you are providing the network with the whole point-cloud that includes both the object to grasp and the surface it rests on? As far as I remember, you should only provide the network with the object point-cloud so the first thing you have to do is to segment out the object from the surface.
You are correct. Been a while since I used this code base so I have forgotten some details.
Maybe the problem is that you are providing the network with the whole point-cloud that includes both the object to grasp and the surface it rests on? As far as I remember, you should only provide the network with the object point-cloud so the first thing you have to do is to segment out the object from the surface.
I have done this ,i remove the plane and only provide the object pc. I have always suspected that my model is not good，but it work well on the test dataset,so i'm very confused