[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
619
stars
76
forks
source link
a little bug for transfusion_head.py, but it does not affect the training results. #54
Many thanks to the authors for open sourcing the TransFusion code. I noticed that in line 827 of the "https://github.com/XuyangBai/TransFusion/blob/master/mmdet3d/models/dense_heads/transfusion_head.py" file, the author has mistakenly written self.img_feat_collapsed_pos as self.img_feat_collapsed_po due to an oversight. But it does not affect the training results, because it is equivalent to recalculating img_feat_collapsed_pos each time.
Many thanks to the authors for open sourcing the TransFusion code. I noticed that in line 827 of the "https://github.com/XuyangBai/TransFusion/blob/master/mmdet3d/models/dense_heads/transfusion_head.py" file, the author has mistakenly written self.img_feat_collapsed_pos as self.img_feat_collapsed_po due to an oversight. But it does not affect the training results, because it is equivalent to recalculating img_feat_collapsed_pos each time.