Reasoning Intention-Oriented Objects (RIO)

RIO数据集由北京交通大学信息科学研究所创建，专注于开放环境中的意图导向对象推理。该数据集包含40,214张图像和130,585对意图-对象标注，旨在通过自然语句描述来识别与特定意图相关的对象。数据集的创建过程涉及使用ChatGPT生成意图描述，并经过人工审核确保描述的合理性和相关性。RIO数据集的应用领域包括计算机视觉中的意图识别和环境感知，旨在提高模型在复杂场景中理解和响应人类意图的能力。The RIO dataset was created by the Institute of Information Science, Beijing Jiaotong University, focusing on intention-oriented object reasoning in open environments. This dataset contains 40,214 images and 130,585 intention-object annotation pairs, aiming to identify objects associated with specific intentions through natural language descriptions. The dataset creation process involved using ChatGPT to generate intention descriptions, which were then manually reviewed to ensure their rationality and relevance. Application areas of the RIO dataset include intention recognition and environmental perception in computer vision, with the goal of enhancing models' ability to understand and respond to human intentions in complex scenarios.