As described in the paper, the original FCAF3D code's backbone network was substituted with our ConDaFormer, leaving the detection head unchanged. For the CAGroup3D method, the ConDaFormer block can be integrated into the backbone network, following the practice of SphereFormer's object detection code. Currently, there are no plans to release the object detection code.
As described in the paper, the original FCAF3D code's backbone network was substituted with our ConDaFormer, leaving the detection head unchanged. For the CAGroup3D method, the ConDaFormer block can be integrated into the backbone network, following the practice of SphereFormer's object detection code. Currently, there are no plans to release the object detection code.