EQA 问答数据集

EQA 全称 Embodied Question Answering，是一个基于 House3D 的视觉问答数据集。在环境中任意位置的 agent 在得到一个问题后，能够自己在环境中寻找有用的信息并对该问题作出回答。比如：Q: 汽车是什么颜色的？为了回答这个问题，agent 必须首先通过智能导航来探索环境，从第一人称视角收集必要的视觉信息，然后回答问题：橙色。EQA, short for Embodied Question Answering, is a visual question answering dataset built upon House3D. An agent placed at an arbitrary position within the environment can, upon receiving a natural language question, autonomously explore the environment to collect useful information and generate a correct answer to the question. For example, given the question "What color is the car?", the agent must first conduct intelligent navigation to explore the environment, collect necessary visual information from a first-person perspective, and then output the answer: "Orange".

搜集汇总

背景与挑战

背景概述

EQA问答数据集是一个基于House3D环境的具身智能视觉问答数据集，要求智能体通过自主导航收集第一视角视觉信息来回答问题。该数据集大小为839.6KB，发布于四年前，主要应用于具身智能、视觉问答和强化学习领域。