文章目录一、2Dopen-vocabularyobjectdetection的发展和研究现状二、基于大规模外部图像数据集2.1OVR-CNN：Open-VocabularyObjectDetectionUsingCaptions，CVPR20212.2OpenVocabularyObjectDetectionwithPseudoBounding-BoxLabels，ECCV20222.2.1伪标签的生成2.2.2检测模型训练2.3Detic:DetectingTwenty-thousandClassesusingImage-levelSupervision，ECCV20222.4Grounde