For machines, solving such a problem requires a model to recognize unseen actions composed of previously observed verbs and objects, thus requiring so-called compositional generalization ability. To facilitate this research, we propose a novel Zero-Shot Compositional Action Recognition (ZS-CAR) task. ...
2019_Task-Driven Modular Networks for Zero-Shot Compositional Learning 2019_ICCV_组合成新概念_组成成分的推理CompositionalReasoning_任务驱动组合_对象属性对 将学习到的知识, 组合成新的概念. 本文提出任务驱动的模块化体系结构, 用于(compositional reasoning)组成成分的推理和样本高效的学习, 捕获图像和'对象-属...
Compositional Zero-Shot LearningImage classificationMutual BalancingVisual-attribute2024 Elsevier LtdCompositional Zero-Shot Learning (CZSL) aims to recognize ... C Jiang,Q Ye,S Wang,... - 《Pattern Recognition the Journal of the Pattern Recognition Society》 被引量: 0发表: 2024年 Tell me what ...
来源: NVIDIA, 巴伊兰大学 论文名称:A causal view of compositional zero-shot recognition 原文作者:Yuval Atzmon 内容提要 人们很容易识别新的视觉类别,这些视觉类别是已知组件的新组合。因为新组合的长尾支配着分布,所以这种组合泛化能力对于在视觉和语言等现实世界中的学习非常重要。不幸的是,学习系统很难与组成概括...
Zero-shot learning is essential for the solution of many problems in the real world. For example, in computer vision, zero-shot learning is applied to video action recognition13 and image recognition14 tasks, which can be performed on unseen videos and images. Therefore, zero-shot learning has...
IEEE Conf. Computer Vision and Pattern Recognition, Honolulu, USA, 2017, pp. 2432–2443.. Google Scholar [106] X. Y. Chen, C. Zhang, G. S. Lin, and J. Han, Compositional prototype network with multi-view comparision for few-shot point cloud semantic segmentation. 2020. [Online]....
Vision-Language相关(5篇)[1] Troika: Multi-Path Cross-Modal Traction for Compositional Zero-Shot Learning 标题:三驾马车:用于组合式零样本学习的多路径交叉模式牵引 链接:arxiv.org/abs/2303.1523 代码:未开源 [2] Multi-Modal Few-Shot Temporal Action Detection 标题:多模式少镜头时间动作检测 链接:arxiv...