S-CLIP Semi-supervised Vision-Language Learning using Few Specialist Captions

• 论文名称: S-CLIP: Semi-supervised Vision-Language Learning using Few Specialist Captions
• 模型架构: VLM
• Visual Encoder: Transformer
• Text Encoder: Transformer
• Model Details: (CLIP)
• Task: Scene Classification, Image-text Retrieval
• Link: https://arxiv.org/abs/2305.14095
• Code/Project: https://github.com/alinlab/s-clip
• Survey Inclusion: awesome-remote-sensing-vision-language-models
• Short Summary: 提出了一种半监督学习方法，用于在少量专家标注的字幕情况下训练CLIP模型。S-CLIP利用额外的未配对图像，并通过两种伪标签策略增强对比学习和语言模态的训练。
• Published in: NeurIPS 2023