-
Notifications
You must be signed in to change notification settings - Fork 300
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
orientation error using 6D pose estimate to display results in 3D #146
Comments
I fixed the issue by changing the callback to
the code here helped. Thanks @hoangcuongbk80 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
@j96w in #88, you mentioned that the canonical frame is the object mesh. If I load the
points.xyz
file into RVIZ, using PCL, the frame appears like thisfor the mustard bottle. So, when the 6D estimate is all zeros, then the canonical frame is equal to the camera frame. But, when the cannonical frame needs to be shifted, I am seeing errors when I use the 6D pose estimate to show the result in 3D
where the RGB pose visualization estimate looks good
Then I publish the 6D pose of the object and use it to transform the
points.xyz
file from the camera frame to the 6D pose for visualization in RVIZ, then the pose estimate looks badAs you can see, it is flipped and twisted, so the orientation is off
FYI: I have tried it several ways, including publishing the tf from the camera to the object then publishing the cloud in the object frame, and they have all given the same result.
First, I was thinking that I need to determine the initial pose of the object, by finding the centeriod of the PLY file and shifting the cloud using that result, similar to #88 yuxng/YCB_Video_toolbox#11 and yuxng/PoseCNN#103 , but I switched to directly using the
points.xyz
file,without shifting and there is no difference.
Might this be due to differences in coordinate systems in RVIZ and the
points.xyz
file? For instance if I flip the z when I import thepoints.xyz
file I getIn NVlabs/Deep_Object_Pose#101 @mintar mentions that the pose topic can be used for visualization, which looks like
But I would like to see the point clouds line up as this pose arrow does not help me evaluate accuracy of the orientation
@hoangcuongbk80 @j96w or @jontremblay or @yuxng or @MrLuer or @mintar any help with this question would be appreciated
The text was updated successfully, but these errors were encountered: