FIGCON: Exploiting FIne-Grained CONstructs of Facial Expressions for Efficient and Accurate Estimation of In-Vehicle Drivers’ Statistics

Zhuoran Bi, Xiaoxing Ming, Junyu Liu, Xiangjun Peng, Wangkai Jin

Designed and proposed FIGCON, a new method to overcome the computation resource constraint and improve existing estimation for in-vehicle drivers’ statistcs using facial expressions. The key idea is to exploit facial expressions into a finer granularity. Our results show that FIGCON can improve the estimation accuracy in most cases, with performance benefits significantly. The results still breed some tradeoffs between the performance and the accuracy for the in-vehicle statistics prediction. We believe our work serve as the starting point for future investigations, to exploit more formalization and exploitation, to take advantage of the fine-grained facial features, so that the accuracy of these estimators.