MDNS

人类可以在一张脸中感知多种表情。这种表达也可以在不同的强度下被感知。通过耦合来自多个观察者的注释，可以观察到某些表达式也是模棱两可的。MDNS (多维，细微差别，主观) 注释数据集为1000面部图像提供了一组新的多维调制注释。这些图像来自标准ExpW人脸图像数据集。这个新颖的野外，种族多样的数据集已由多个众包注释器沿6、15和21个表达维度进行注释。在这里，我们提供了原始注释器评级，并演示了一种可用于模拟人类感知的概率技术。Humans can perceive multiple expressions in a single face, and these expressions can also be perceived at varying intensities. By coupling annotations from multiple observers, it can be observed that certain expressions are ambiguous. The MDNS (Multi-dimensional, Nuanced, Subjective) annotation dataset provides a new set of multi-dimensionally modulated annotations for 1000 facial images sourced from the standard ExpW facial image dataset. This novel in-the-wild, ethnically diverse dataset has been annotated by multiple crowdsourced annotators along 6, 15, and 21 expression dimensions. Here, we provide the raw annotator ratings and demonstrate a probabilistic technique that can be used to model human perception.

搜集汇总

背景与挑战

背景概述

MDNS是一个多维、细微差别、主观的人脸表情注释数据集，包含1000张来自ExpW数据集的图像，由多个众包标注者沿6、15和21个表达维度进行注释，并提供原始评级及概率技术来模拟人类感知。该数据集由佐治亚理工学院和AWS AI Labs于2022年发布。