-
Notifications
You must be signed in to change notification settings - Fork 1
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
Wrong rotation angle of converted 3D bounding box labels #1
Comments
Issue #1 displays a skew in the yaw angle of the 3D labels. Modifications will be made.
Hello @kyrie-louy. Thank you for bring this problem to my attention. I've just resolved it and have pushed the updated script. Now, try the conversion again, and it should be accurate. |
Hi @ZiadElmassik. Thank you for your prompt reply! The problem for the 3D bounding box is resolved, while the corresponding 2D bounding boxes went wrong: I'm also trying to figure out this issue.
|
Alright no worries @kyrie-louy, I'll work on it and respond to you with a solution in a few hours. |
Alright @kyrie-louy, the 2D bounding boxes are fixed along with the 3D rotations of the 3D bounding boxes. I just pushed the updated script. Please, try the conversion again. |
Yeah, I see the reason, and everything looks fine now. Many thanks for providing the solution! |
Hi @ZiadElmassik, thanks for your great work. It helps me a lot!
However, after I converted CADCD to KITTI and visualized it, I found that the rotation angle of the ground truth 3D bounding box for many objects is incorrect. And this issue exists in each frame (both 2018 and 2019 samples).
Here I select the frame
cadcd/2018_03_06/0001/labeled/image_00/data/0000000045.png
for your reference. We can see that the GT bbox of the front car, as well as the GT bbox of vehicles on the right side, have very strange angles.3D view:
BEV view:
Camera view:
I wonder that have you encountered such an issue? And what may be the possible reason?
Note that I follow the instruction in README to comment, and I also repeat the visualization through the tools in second.pytorch and kitti_obj_vis.
The text was updated successfully, but these errors were encountered: