Hello Nvidia Developer Forum Administrator,
I would like to inquire about the category of 3D Object Detection Models.
Aside from PointPillars, are there currently any other 3D Object Detection Models in development or progressing ?
Thank you!
There is BEVFusion which is an 3D object-detection model that is included in the TAO. Please take a look at doc
BEVFusion - NVIDIA Docs and notebook tao_tutorials/notebooks/tao_launcher_starter_kit/bevfusion at main · NVIDIA/tao_tutorials · GitHub.
1 Like
Thank you!
I’d like to ask if there are any recent updates on 3D object detection models that are voxel-based, specifically those using only raw point cloud data as input (excluding RGB images or fusion-based models like BEVFusion).
There is not updated on this kind of model. Only PointPillars is for end users.
An update here. We have centerpoint internally which is similar to your request.
1 Like
@Morganh
Great job to the entire Nvidia team!
Looking forward to any updates or releases related to Centerpoint—please keep us posted!
By the way, when I trained the PointPillars model, I noticed the label text file in the infer folder includes lines like:
Is the last column the probability of detection for the object?
57735747-271516919_604653692-414852285_43749964-86842239.txt (314 Bytes)
Could you please create a new topic for this? Thanks a lot!
1 Like
Currently, Centerpoint is not in the release roadmap of TAO. It is just from one personal project of developer team. Could you please contact the sales team of Nvidia? Then the sales team can contact the developer team about the release process. Thanks.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.