A Demonstration of KAMEL: A Scalable BERT-based System for Trajectory Imputation

This demo presents KAMEL; a novel trajectory imputation framework that aims to impute sparse trajectories as a means of increasing their accuracy, and hence the accuracy of their applications. Unlike the large majority of current trajectory imputation techniques, KAMEL does not require the knowledge or the availability of the underlying road network, which makes it applicable to important applications like map inference that need to infer the road network itself. Audience will experience KAMEL through various scenarios that show the imputation accuracy as well as KAMEL internals.