The second stage of training with pdv is not convergent (training end loss~=9.0), is there any detail that needs attention?

PJLab-ADG / DetZero

[ICCV 2023] DetZero: Rethinking Offboard 3D Object Detection with Long-term Sequential Point Clouds

https://superkoma.github.io/detzero-page/

Apache License 2.0

334 stars 32 forks source link

The second stage of training with pdv is not convergent (training end loss~=9.0), is there any detail that needs attention? #33

Open Winston-Lii opened 1 year ago

Winston-Lii commented 1 year ago

In addition, the result of inference using pdv model cannot track the object in tracking. Is it possible that this is related to the problem described in the title @