【論文要約:自動運転関連】Make Your ViT-based Multi-view 3D Detectors Faster via Token Compression