方法对比

	多模态Transformer	Vision Transformer
领域	深度学习	深度学习
方法族	Machine learning	Machine learning
起源年份	2019–2021	2021
提出者	Lu et al. (ViLBERT); Radford et al. (CLIP)	Dosovitskiy, A. et al.
类型	Cross-modal attention-based deep learning model	Transformer architecture for images (self-attention over patches)
开创性文献	Lu, J., Batra, D., Parikh, D., & Lee, S. (2019). ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks. Advances in Neural Information Processing Systems (NeurIPS), 32.	Dosovitskiy, A. et al. (2021). An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. ICLR.
别名	multimodal attention model, cross-modal transformer, vision-language transformer, multi-modal fusion transformer	Görsel Transformer (ViT), görsel transformer, ViT, patch transformer for images
相关	5	5
摘要	A Multimodal Transformer extends the standard Transformer architecture to process and jointly reason over two or more input modalities — most commonly text and images, but also audio, video, or structured data. Cross-modal attention layers allow information from one modality to inform representations in another, enabling tasks such as visual question answering, image captioning, and multimodal sentiment analysis.	The Vision Transformer (ViT), introduced by Dosovitskiy and colleagues in 2021, splits an image into fixed-size patches, treats those patches as a sequence, and applies the Transformer self-attention mechanism to image classification. Given enough training data, it surpasses convolutional neural networks (CNNs).