Konyul Park

Question about submission

2

I want to know whether this paper is accepted to ECCV 2022 thank you.

Inference speed

3

When I check the inference speed of Centerpoint(pointpillars,voxelnet) I get the following FPS (2 task/s, 4.7 task/s).  command I typed is python tools/dist_test.py configs/nusc/pp/nusc_centerpoint_pp_02voxel_two_pfn_10sweep.py --checkpoint nusc_centerpoint_pp_02voxel_two_pfn_10sweep_latest.pth --work_dir nusc_centerpoint_pp_02voxel_two_pfn_10sweep --speed_test...

Double flip

5

Do we have to train model with double flip to enhance mAP? Because when I test model with and without doble flip, model with doble flip get worse mAP.

Hi, I want to assign pillar indexes after downsampling into pillar indexes before downsampling. I used pair_bwd to solve this problem, but The pillar indexes after downsampling were not fully...

learning rate configuration

2

Thank you for sharing such a great work! I am struggling to reproduce the L+C result of Bevfusion. Currently, the repreduced result is 70.14 NDS. What I did was modifying...

 Can you provide the config file that you used to generate the result PillarNet-18(neckv1)??

Question about selecting DAM as ground truth of salient token prediction

1

why didn't you set the real ground truth as ground truth of salient token prediction??

The paper explains about implementation details that the weights of the image backbone network was frozen during training. So I wonder how the performance would be if the weights of...

VoxelNeXT-2D

1

can you share pretrained model for voxelnext-2d on nuscenes?

Hi, thank you for sharing such a great work. Currently, I understand that BEVFusion is trained in a 2-stage process. In the 1st stage, following Transfusion-L, the LiDAR branch is...