about two stage centerpoint

[fukexue]

I'm sorry to bother you.
I'd like to ask if the idea of 4.1 two stage centerpoint in this article has not been implemented in centerpoint Kitti. Because I didn't find them in the repo. We look forward to your reply.

The article idea as follow:
We extract one point-feature from the 3D center of each face of the predicted bounding box. Note that the bounding box center, top and bottom face centers all project to the same point in map-view. We thus only consider the four outward-facing box-faces together with the predicted object center. For each point, we extract a feature using bilinear interpolation from the backbone map-view output M. Next, we concatenate the extracted point-features and pass them through an MLP. The second stage predicts a class-agnostic confidence score and box refinement on top of one-stage CenterPoint's prediction results.

[tianweiy]

Hi, yes, this repo only includes the first part. you can refer to this repo https://github.com/tianweiy/CenterPoint/blob/master/det3d/models/detectors/two_stage.py to see how we implement two stage on Waymo.

I tried this on KITTI a while ago and it is not much better than the first stage one so I didn't implement it here. I would suggest combining with other two stage approaches like PVRCNN for the best result.