CLIP2Scene: Towards Label-Efficient 3D Scene Understanding by CLIP | Read Paper on Bytez