Masking the Truth: Effective facial recognition in a COVID-19 world

掩盖真相：COVID-19 世界中的有效面部识别

• 批准号: 83956
• 金额: $ 7.26万
• 依托单位: TUMBLING DICE LIMITED
• 依托单位国家: 英国
• 项目类别: Collaborative R&D
• 财政年份: 2020
• 资助国家: 英国
• 起止时间: 2020 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=83956
• 关键词: Masking; Truth; Effective; facial; recognition

The aim of this project is to develop a new facial recognition system, one which can perform accurately and reliably on partially occluded faces. The need for this has arisen as a direct result of the COVID-19 pandemic, where facial coverings have not only become common place but in many instances are now compulsory.In July 2020, the US National Institute of Standards and Technology (NIST) published a report which stated that facial recognition algorithms developed before the emergence of COVID-19 have "_great difficulty_" in accurately identifying people wearing facial coverings.According to NIST: "_Even the best of the 89 commercial facial recognition algorithms tested had error rates between 5% and 50% in matching digitally applied face masks with photos of the same person without a mask._"During this project we will implement a facial recognition system which is;* Generic. Meaning it will be capable of efficiently classifying any image.* Can be run on low power hardware. Meaning it will be energy efficient and limit GHG (greenhouse gas) emissions.* Capable of self-directed dynamic learning. Meaning it will be able to learn 'on the job'.* Shows no subject bias.A recognition system with these attributes will be disruptive, game changing and a clear advance over the state-of-the-art.

该项目的目的是开发一种新的面部识别系统，该系统可以在部分遮挡的面部上准确可靠地执行。COVID-19大流行直接导致了这种需求的产生，面部覆盖物不仅成为常见的地方，而且在许多情况下现在是强制性的。美国国家标准与技术研究所（NIST）发表了一份报告，称在COVID-19出现之前开发的面部识别算法具有"极大的困难"，据NIST称：“即使是测试的89种商业面部识别算法中最好的算法，在将数字应用的面罩与同一个人不戴面罩的照片进行匹配时，错误率也在5%到50%之间。"在这个项目中，我们将实现一个面部识别系统，这是;* 通用。这意味着它将能够有效地对任何图像进行分类。*可以在低功耗硬件上运行。这意味着它将是能源效率和限制GHG（温室气体）排放。能够自我导向的动态学习。这意味着它将能够“在工作中”学习。具有这些属性的识别系统将是破坏性的，改变游戏规则的，并且是对最先进技术的明显进步。