ecoset

Ecoset是一种生态动机的图像数据集，是为人类视觉神经科学设计的大规模图像数据集，由来自565基本级别类别的150万多个图像组成。类别选择基于最常以口语出现的英语名词 (根据从美国电视和电影字幕中获得的一组5100万单词估算) 和人类观察者的具体等级。Ecoset由基本级别的类别 (包括人类类别的男人，女人和孩子) 组成，这些类别描述了对人类重要的世界上的物理事物 (而不是抽象的概念)。

Ecoset is an ecologically motivated image dataset designed for human visual neuroscience research. It is a large-scale image dataset comprising over 1.5 million images from 565 basic-level categories. The selection of its categories is based on two aspects: the most frequently occurring spoken English nouns (estimated from a corpus of 51 million words collected from American television and film subtitles) and the concrete ratings from human observers. Ecoset consists of basic-level categories—including the human-related categories of man, woman, and child—that depict physical entities in the world that are significant to humans, rather than abstract concepts.

搜集汇总

背景与挑战

背景概述

Ecoset是一个专为人类视觉神经科学研究设计的大规模图像数据集，包含超过150万张图像，覆盖565个基于英语常用名词和人类观察者具体等级选择的基本级别类别，这些类别主要描述对人类重要的物理实体。该数据集由剑桥大学、哥伦比亚大学和Radboud University于2021年联合发布。