Simulation experiment

[sazanali]

Dear Sir,
I read your work “ Rethinking IoU-based Optimization for Single-stage 3D Object Detection”. It was great and I have questions about the simulation work you did in this work, you speak about simulation experiment for two specific anchor size and you measured the gradient, the iou and RDIoU values with angle . Is it possible that I can see this simulation experiment with its code?, could you help me and send me this simulation experiment code?

Regards

[NiranjanRavi1993]

Hi,
Were you able to find the source repo for the simulation experiment?
Regards
Niranjan.

[hlsheng1]

Hi, the source code about this simulation part is erased in a machine updating (preparing for Foundation Models leanring). I remember this part is done by firstly saving the grad of each parameter by loss backward step in one iteration, and then ploting them.

[NiranjanRavi1993]

Hi @hlsheng1 Thank you for the prompt reply. I will try your suggestion. I had an additional question regarding the checkpoint for "car" class. To obtain this checkpoint, how many epochs were trained? 62? It would be helpful to know.

regards
Niranjan Ravi.