sensor2sensor_mats doesn't seem to be used.

[Camellia-hz]

@GOATmessi7 hello, May I ask how the feature map of keyframes and non-keyframes are aligned during the training process? Is the parameter sensor2sensor_mats used? But I don't use the parameter sensor2sensor_mats in BEVDepth/bevdepth/layers/backbones/base_lss_fpn.py file.

Is it appropriate to directly generate and ice a bev feature map of keyframes and non-keyframes like the code in the image?

[Camellia-hz]

@tonysy