Visual Recipe Flow
收藏arXiv2022-09-13 更新2024-08-06 收录
下载链接：
http://arxiv.org/abs/2209.05840v1
下载链接
链接失效反馈官方服务：
资源简介：
Visual Recipe Flow是一个多模态数据集，由京都大学创建，专注于烹饪过程中的物体状态变化。该数据集包含3705个物体的状态变化，每个变化通过一对图像表示，并与烹饪流程图（rFG）关联，以支持跨模态推理。数据集的创建涉及详细的图像和文本标注，旨在帮助开发能够理解和预测烹饪动作结果的自主代理。该数据集适用于多模态常识推理和程序文本生成等应用，以提高对烹饪过程的理解和自动化。
Visual Recipe Flow is a multimodal dataset developed by Kyoto University, focusing on object state changes during cooking processes. This dataset contains 3705 object state changes, each of which is represented by a pair of images and linked to recipe flow graphs (rFG) to support cross-modal reasoning. The construction of this dataset involves detailed image and text annotations, aiming to assist in developing autonomous agents that can understand and predict the outcomes of cooking actions. This dataset is suitable for applications such as multimodal commonsense reasoning and procedural text generation, to improve the understanding and automation of cooking processes.
提供机构：
京都大学
创建时间：
2022-09-13
