[PyTorch] Official implementation of CVPR2022 paper "TransFusion: Robust LiDAR-Camera Fusion for 3D Object Detection with Transformers". https://arxiv.org/abs/2203.11496
Thanks for the interesting research.
I understand filtering the points that are occluded by objects in the GT-sampling step. However, in GT sampling, objects can be created in unexpected spaces. Is there a way to get rid of GT sampled objects if they are created behind walls or behind non-annotated objects? In lidar image fusion, this situation seems to hinder learning.
Thanks for the interesting research. I understand filtering the points that are occluded by objects in the GT-sampling step. However, in GT sampling, objects can be created in unexpected spaces. Is there a way to get rid of GT sampled objects if they are created behind walls or behind non-annotated objects? In lidar image fusion, this situation seems to hinder learning.
thank you Minseok