Fine-Grained Image-Text Correspondence with Cost Aggregation for Open-Vocabulary Part Segmentation | Read Paper on Bytez