02-04 ECCV2022 A Simple Baseline for Open-Vocabulary Semantic Segmentation with Pre-trained Vision-language Model