Finn

The only thing I found related to transformation of prev_bev is [here](https://github.com/fundamentalvision/BEVFormer/blob/32b65ffd2a85c35b5f74b753f23ef34f9a85a576/projects/mmdet3d_plugin/bevformer/modules/transformerV2.py#L143C29-L143C29) However, I think this part is related to data augmentation only

> In v2, Its implemented by align the cam2lidar transformation matrix, maybe called online camera on many other framework. See here > > https://github.com/fundamentalvision/BEVFormer/blob/32b65ffd2a85c35b5f74b753f23ef34f9a85a576/projects/mmdet3d_plugin/datasets/nuscenes_dataset_v2.py#L137 > > > In v1, Its...

It seems that this alignment method is the same with this paper [TBP-Former](https://arxiv.org/abs/2303.09998)

Thanks for your reply! Here I also have a silly question. In Nerfacc we have something like this ` weights = render_weight_from_alpha(alpha, ray_indices=ray_indices, n_rays=n_rays)` ` opacity = accumulate_along_rays(weights, ray_indices, values=None,...

Thanks for your reply and help! Appreciate!!!