ZegCLIP: Towards Adapting CLIP for Zero-Shot Semantic Segmentation | Read Paper on Bytez