search

XuyangBai / TransFusion

[PyTorch] Official implementation of CVPR2022 paper "TransFusion: Robust LiDAR-Camera Fusion for 3D Object Detection with Transformers". https://arxiv.org/abs/2203.11496
Apache License 2.0
642 stars 77 forks source link

Transfusion model with encoders #23

Open diegodibe opened 2 years ago

diegodibe commented 2 years ago

Dear authors, thank you for your work. I wonder weather it has been tested a version of the transformer with encoder layers. Or if there is any reason of why the encoders are not used. Thank you

XuyangBai commented 2 years ago

Hi, personally the encoder layers of the transformer are for enhancing the original feature, where I do not think it is necessary for our case since the lidar feature extracted by the 3DCNN backbone are already good. The brought improvement of adding an encoder might not be attractive given the increased computation cost.