CgT-GAN: CLIP-guided text GAN for image captioning

• Main Authors: YU, Jiarui, LI, Haoran, HAO, Yanbin, ZHU, Bin, XU, Tong, HE, Xiangnan
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2023
• Subjects: Image captioning; CLIP; Reinforcement learning; GAN; Graphics and Human Computer Interfaces; Programming Languages and Compilers
• Online Access: https://ink.library.smu.edu.sg/sis_research/9012; https://ink.library.smu.edu.sg/context/sis_research/article/10015/viewcontent/CgT_GAN.pdf
• Institution: Singapore Management University

The large-scale visual-language pre-trained model, Contrastive Language-Image Pre-training (CLIP), has significantly improved image captioning for scenarios without human-annotated image-caption pairs. Recent advanced CLIP-based image captioning without human annotations follows a text-only training...