Crowd4ME: A Crowdsourcing-Based Micro-expression Collection Platform

Abstract

Machine learning-based approaches have greatly accelerated the progresses in the field of micro-expression detection and recognition. However, many models, especially those based on deep learning require very large databases of hand labeled data for training. Existing micro-expression data collection processes usually cost a lot and are time-consuming. The labeling of very large datasets is becoming a bottleneck of building robust models for micro-expression detection and recognition. With the wide success of crowdsourcing platforms in creating large datasets like ImageNet, in this paper, we present Crowd4ME. Crowd4ME is a crowdsourcing-based platform for collecting large scale micro-expression data. We demonstrate that using Crowd4ME can help collect and manage micro expression samples more easily and efficiently.