However, inaccurate depth estimation and substantial semantic information loss remain significant factors limiting the accuracy of 3D detection. In this paper, we propose a cross-fusion framework ...