Search for a command to run...
Learning to Generate Text-grounded Mask for Open-world Semantic Segmentation from Only Image-Text Pairs