yichen928 / SparseFusion

[ICCV 2023] SparseFusion: Fusing Multi-Modal Sparse Representations for Multi-Sensor 3D Object Detection
Apache License 2.0
198 stars 19 forks source link

About the LiDAR branch #18

Closed qzhang852 closed 1 year ago

qzhang852 commented 1 year ago

Dear author, Thank you for your excellent work. I wonder how do you get LiDAR branch with semantic transfer, add the

sample_query_feature = sample_query_feature + sample_query_img_feat

directly in the transfusion_nusc_voxel_L.py ? Thank you for taking the time.

yichen928 commented 1 year ago

Thanks for your interest. This line is in the semantic transfer module. This module gives a better initial query position for the LiDAR branch, but the initial query features of LiDAR branch still come from only the LiDAR branch. In this case, we can directly re-use the architecture and pretrained weight of transfusion-L without any modifications.