SCI-CLIP: Segment-Centric Inference with Reference Memory for Training-Free Open-Vocabulary Segmentation figure
AlphaXiv 中文概览(可滚动查看)