CLIP as RNN: Segment Countless Visual Concepts without Training Endeavor | Read Paper on Bytez