方法对比

	多模态实例分割	多模态视觉变换器
领域	深度学习	深度学习
方法族	Machine learning	Machine learning
起源年份	2017–present	2021
提出者	He, K., Gkioxari, G., Dollar, P., Girshick, R. (Mask R-CNN foundation); extended by community to multimodal settings	Dosovitskiy et al. (ViT); Radford et al. (CLIP multimodal ViT)
类型	Supervised deep learning — instance segmentation	Multimodal transformer model
开创性文献	He, K., Gkioxari, G., Dollar, P., & Girshick, R. (2017). Mask R-CNN. Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2961–2969.	Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., & Houlsby, N. (2021). An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. In International Conference on Learning Representations (ICLR).
别名	multimodal Mask R-CNN, RGB-D instance segmentation, multi-sensor instance segmentation, cross-modal instance segmentation	Multimodal ViT, vision-language transformer, cross-modal vision transformer, multi-modal ViT
相关	5	5
摘要	Multimodal instance segmentation extends classical instance segmentation — which assigns a per-pixel mask and a class label to every individual object in an image — by incorporating complementary sensor streams such as depth maps, LiDAR point clouds, or infrared frames. Fusing these modalities helps the model handle ambiguous appearances, low light, and occlusion that trip up RGB-only systems.	Multimodal Vision Transformer (Multimodal ViT) extends the Vision Transformer architecture to jointly process and align representations from multiple modalities — typically images and text — using self-attention and cross-attention mechanisms. By learning shared or aligned embedding spaces across modalities, it enables tasks such as visual question answering, image-text retrieval, visual grounding, and image captioning.