search

SerCharles / CN-RMA

Official implementation of CN-RMA: Combined Network with Ray Marching Aggregation for 3D Indoor Object Detection from Multi-view Images
Other
15 stars 2 forks source link

Question about nerfdet results reported in your paper #1

Closed Pixie8888 closed 6 months ago

Pixie8888 commented 6 months ago

Hi author,

I found that the nerfdet results reported in your paper (table 1 and table 2) are inconsistent with any existing papers. How did you get these results?

SerCharles commented 6 months ago

I directly run the code on the NeRFDet paper to get the ScanNet result using their provided checkpoint. Since it does not provide official code and checkpoint on the ARKitScenes dataset, I slightly change the config, change the bbox_head to bbox_head=dict( type='SunRgbdImVoxelHeadV2', n_classes=17, n_channels=128, n_reg_outs=7, n_scales=3, limit=27, centerness_topk=18), following ImVoxelNet, which use the head on the OBB dataset of SUNRGBD, and get my result. This result is slightly higher than NeRFDet paper provided.

Pixie8888 commented 6 months ago

thanks!